The Islamist Advantage: The Religious Infrastructure of Electoral Victory

Sharan Grewal∗ March 29, 2021

Abstract

Why do Islamists regularly win elections in the Middle East? One common yet

rarely tested hypothesis is that Islamists can rely on a vast, country-wide network of

religious institutions, particularly , to facilitate voter outreach and mobiliza-

tion. Secular parties, meanwhile, have no comparable, preexisting infrastructure. This

paper attempts to test this “infrastructure advantage” in two ways. First, it leverages a

dataset of ’s 6000 mosques to show that sub-nationally, the number of mosques

per capita strongly correlates with Islamist vote share in the 2011, 2014, and 2019

parliamentary elections. Notably, results appear to be driven by mosques facilitating

personal, horizontal interactions with Islamists, rather than exposure to Islamist social

services, politicized imams, or underlying religiosity. Second, drawing on region-wide

survey data from the Arab Barometer, it shows that respondents who attend

for Friday prayers are significantly more likely to trust Islamists, and that this is one

of the most consistent predictors of across multiple survey waves.

∗Assistant Professor, William & Mary ([email protected]). I am indebted to the phe- nomenal research assistance of Cassie Heyman-Schrum, Hank Hermens, and Amy Hilla. For helpful comments, I thank Nasir Almasri, Lucia Ardovini, Elizabeth Baisley, Steven Brooke, Melani Cammett, Matthew Cebul, M.Tahir Kilavuz, Alex Kustov, Marc Lynch, Ameni Mehrez, Elizabeth Nugent, Tom Pavone, Scott Williamson, and audiences at POMEPS, Oxford, Tulane, and Central European University. 1 Introduction

When Islamists run in competitive elections in the Middle East, they tend to win. Beginning with the Islamic Republican Party’s victory in in the wake of the Islamic Revolution, the region has seemingly witnessed a ‘green wave,’ with Islamists racking up electoral victories in Algeria, , Palestine, Iraq, Egypt, Tunisia, and Morocco, among others. Even in the past three years, Islamists have placed first in all three of the competitive elections in the region, namely Lebanon (2018), Iraq (2018), and Tunisia (2019). In total, Islamists have won 20 out of the 30 competitive elections (67%) they have contested in the Middle East since the 1970s (see figure1). Naturally, a large literature has emerged attempting to explain this ‘Islamist advantage’ in elections. For some, their victory reflects the region’s widespread embrace of Islamist ideologies, especially as a reaction to colonialism and forced secularization (Ayubi, 1991; Esposito, 1997; Nugent, Masoud and Jamal, 2018). For others, the appeal of Islamists is their services, whether the charity and welfare they provide to would-be voters (Wedeen, 2003; Cammett and Issar, 2010), the expectation that they will redistribute wealth once in office (Masoud, 2014), or the divine rewards they promise in the afterlife (Grewal et al., 2019). Finally, others argue that the Islamist advantage lies instead in their reputations for competence, probity, and authenticity (Cammett and Luong, 2014; Brooke, 2019). However, there is also an earlier literature claiming that Islamists are not necessarily more appealing than secular parties, but instead that they are better resourced (Wickham, 2002; Clark, 2004; Wiktorowicz, 2004). Social movement theorists, in particular, have highlighted that Islamists enjoy what we might term an “infrastructure advantage.” In this account, Islamists benefit from a nation-wide network of religious institutions, particularly mosques, that facilitate their outreach to new supporters. Secular parties, meanwhile, have no such preexisting infrastructure to aid voter outreach and mobilization. While intuitive, this ‘infrastructure advantage’ has rarely been subject to empirical anal- yses. Masoud(2014) is the lone exception, who finds only mixed evidence that mosques

1 provided Islamists an advantage in elections in Egypt. Outside of the Middle East, Bazzi, Koehler-Derrick and Marx(2020) argue that religious institutions encouraged the rise of Islamists in Indonesia, while Jamal(2005) and Moutselos(2020) find that mosque atten- dance increases civic engagement and voter turnout among Muslims in the US and Europe. However, whether and to what extent mosques help explain the Islamist advantage in the Middle East remains an open question. In this paper, I provide two empirical tests of Islamists’ infrastructure advantage in the Middle East. First, I leverage an original dataset of all 6000 mosques in Tunisia to conduct an ecological analysis of mosque density and Islamist vote share. I show that the number of mosques per capita at the delegation level strongly correlates with the Islamist party Ennahda’s vote share in the 2011, 2014, and 2019 parliamentary elections. Moreover, mosque density also correlates with the 2019 vote share of a new Islamist party, the Karama Coalition, as well as the total vote share of all Islamist parties combined. However, one possible confounder to this relationship is religiosity, which might be pro- ducing both a demand for mosques and for Islamists. I accordingly control for religiosity in two ways. First, I combine six nationally-representative surveys (four waves of the Arab Barometers and two of my own), to generate a sufficient level of respondents (N=8,203) to make meaningful claims about piety at the delegation level. Second, I follow Ciftci, Robbins and Zaytseva(2020) and collect satellite imagery of night-time lights during as a proxy for sub-national religiosity. The link between mosque density and Islamist vote share remains significant in the presence of either of these proxies, helping to rule out religiosity as a confounder. The results are also robust to a variety of demographic and political covariates, including population density, youth, unemployment, education, urban/rural, and proxies for the former dictatorship’s human rights violations, corruption, and regional discrimination. I then explore the mechanisms in Tunisia by conducting three large-scale surveys (total N=13,000) around the 2019 elections. I find no evidence that the effect of mosque density on Islamist vote share in 2019 was driven by politicized imams exhorting their followers to vote

2 for Islamists, by Islamists campaigning near mosques, or by social service provision. Instead, I privilege a more subtle explanation, that mosques provide Islamists the infrastructure to informally meet new supporters and build connections through horizontal, face-to-face interactions. Finally, to show that this relationship might generalize beyond Tunisia, I explore region- wide survey data from the Arab Barometer. I find that respondents who self-report attending mosque for Friday prayers are significantly more likely to trust Islamists, even when control- ling for religiosity and other explanations of Islamist support. Moreover, mosque attendance is one of the most consistent predictors of Islamism across multiple survey waves. In sum, this paper provides some of the first systematic evidence of the infrastructure advantage: the idea that Islamists win elections in the Middle East because they can rely on a vast network of religious institutions, particularly mosques, to aid voter outreach and mobilization. While recent literature seems to privilege the Islamist appeal in elections, suggesting that their services, reputations, or ideologies attract voters, our results instead suggest that more attention should be paid to the differential resources that Islamist parties enjoy in creating ties with potential voters.

2 The Islamist Advantage

Elections in the Middle East are not often free and fair, but when they are, they tend to be won by Islamists. Figure1 plots the 30 competitive 1 parliamentary elections that Islamist parties have contested in the region, from the National Salvation Party’s lackluster performance in the 1973 Turkish election, to Ennahda’s victory in Tunisia in 2019. The y-

1Competitive coded by National Elections Across Democracy and Autocracy (NELDA) project (https://nelda.co/) as 1) opposition were allowed to run, 2) more than one party was legal, 3) there was a choice of candidates on the ballot, 4) the incumbent did not enjoy media bias, and 5) there were no allegations by Western observers of significant vote fraud.

3 axis presents the total vote share obtained by all Islamist parties contesting those elections, while the color scheme records whether the winning party was Islamist (blue) or not (red).

Figure 1: The Islamist Advantage, 1970-2020

The data show that Islamists won 20 of these 30 parliamentary elections, or 67%. In other words, when Islamists run in competitive elections in the Middle East, they are more likely to win than lose. Moreover, across these 30 elections, Islamists garnered an average of 32% of the vote, ranging from the Welfare Party’s 7% in Turkey in 1987 to the 74% blowout by the Muslim Brotherhood and Salafi Nour party in Egypt’s 2012 Shura Council elections. Given the extent of Islamist domination, scholars have focused extensively on explaining the Islamist advantage in elections, putting forth four primary sets of explanations. First, some explain the victory of Islamist parties as reflecting the widespread embrace of Islamist ideologies. In this account, voters in the Middle East turned to Islamism as a reaction to colonialism, forced secularization, and Western influence (Ayubi, 1991; Esposito, 1997; Kepel, 2002; Wolf, 2017; Nugent, Masoud and Jamal, 2018), seeking instead an ideology with greater cultural resonance and which better accorded with their conservative religious beliefs. The 1967 defeat to Israel in particular is often pinpointed as the moment when secular Pan- Arabism was shattered and Islamism stepped in to fill the ideological void (Haddad, 1992).

4 Pew surveys showing widespread support for shari‘a, including hudud punishments, lend credence to the Islamists’ ideological hegemony (Pew Research Center, 2013).2 Second, others argue that the appeal of Islamists is not their ideology, but rather their services. In this narrative, the state’s retreat from providing social services due to structural adjustment policies in the 1980s and 1990s allowed Islamist charities and clinics to fill in the gaps (Wedeen, 2003), leading them to earn votes either through traditional patron-client networks (Cammett and Issar, 2010; Cammett, 2014), middle class networks (Clark, 2004), or by creating a reputation for good governance (Cammett and Luong, 2014; Brooke, 2017, 2019). More generally, the region’s chronic unemployment, poverty, and underdevelopment are thought to increase support for economic redistribution, a policy which voters (often incorrectly3) attribute to Islamists (Pepinsky, Liddle and Mujani, 2012; Masoud, 2014).4 Finally, others argue that it is not their material services in this life, but their promise of divine rewards in the afterlife that attracts the economically strained (Grewal et al., 2019). A third set of arguments locates the Islamist advantage not in their ideologies or their services, but instead in their reputations. Islamists are thought to carry a reputation for being uniquely “competent, trustworthy, and pure,” due in part to their piety and effective service provision (Cammett and Luong, 2014, p. 199). As Islamist leaders often hail from similar socioeconomic backgrounds as their constituents, they are viewed as more authentic and ‘of the people’ (Wickham, 2002) than their typically more elite secular counterparts. Others claim that their time in the opposition under dictatorship gave Islamists a reputation for being serious and effective advocates for political reform (Shehata and Stacher, 2006).

2However, note also Yildirim and Lancaster(2015)’s finding that more moderate Islamist ideologies tend to perform better in elections. 3Once in power, Islamists often pursue neoliberal economic reforms. See al Anani(2020) for Egypt’s Muslim Brotherhood and Ben Salem(2020) for Tunisia’s Ennahda. 4Pellicer and Wegner(2014), however, find that literacy and education are positively, not negatively, associated with support for Morocco’s PJD.

5 Finally, Islamists may be personally more credible in their promises to reform the system due to their history as victims of repression, a status which signals their commitment to the cause (Guiler, 2020). Each of these reputations – credible, competent, authentic, incorruptible – are likewise thought to fuel support for Islamists. While each of the previous three arguments highlight something appealing or attractive about Islamists, a fourth and final argument instead emphasizes Islamists’ superior orga- nization. For some, Islamists command exceptionally disciplined and obedient followers, who turn out to vote more reliably than supporters of other parties (Kandil, 2015; Trager, 2016). For others, Islamists benefit from the plethora of preexisting religious institutions, which give them an edge over secular parties in voter outreach and mobilization (Wickham, 2002; Clark, 2004; Wiktorowicz, 2004). Of particular focus is the mosque, which Islamists can use to preach their message. For instance, in explaining the victory of Algeria’s Islamic Salvation Front (FIS) in 1991, Willis(1997, p. 122) observes that: “Full use was made of the network of some 9,000 mosques which Islamists had come to control, to coordinate activity and spread the party’s message at the weekly Friday prayers and sermon.”5 However, the importance of the mosque has also been contested by other scholars, who argue that regimes across the region – at least after the FIS’ victory – have tightly controlled mosques to prevent Islamists from using them in this fashion (Cammett and Luong, 2014, p. 196). Governments closely regulate who can become an imam, and often even regulate the sermons they can deliver (Wiktorowicz, 1998). Other regimes close down mosques in between prayers to prevent Islamists from leading informal study groups or teaching religious lessons. Accordingly, many argue that the mosque advantage may be overstated. In light of this challenge, this paper seeks to revive the importance of the mosque, in particular by highlighting how mosques might lend Islamists an advantage even when they are not preaching or teaching from the top-down.

5For the importance of the mosque in other cases, see Abu-Amr(1993, p. 8), Waltz (1986, p. 656), Boulby(1988, p. 602), and Wolf(2017, p. 36).

6 3 The Infrastructure Advantage

Even under conditions of repression and surveillance, when Islamists may not be in “control” of a mosque, the mosque still provides a unique opportunity for Islamists to expand their social networks. Put simply, even when Islamists are attending, not preaching, at the mosque, the mosque still provides a forum for meeting other attendees. Two aspects of the mosque facilitate these social interactions. First, mosques bring together the relatively more religious subset of society. While of course not all pious Muslims are Islamists, religiosity does on average correlate with support for Islamism (i.e., Tessler, 2010; Gidengil and Karakoc, 2016). The mosque thus assembles a generally more sympathetic audience for Islamists to meet than in other fora. Second, the mosque offers regular and repeated interactions with other mosque-goers. prescribes five daily prayers, and encourages believers to perform them at a mosque. Even for Muslims who only attend mosque for the large, Friday prayers (Sala¯t al-Jumu‘ah), the mosque still offers an opportunity for Islamists to meet them weekly. What begins as casual interactions in the mosque can over time develop into much stronger social bonds. Dina Zakaria, a spokeswoman for the Egyptian Muslim Brotherhood, observes that such interactions in the mosque were her introduction to the movement:

“From my own experience, those who are not related to the Muslim Brotherhood by family, get to know them in life, such as in the mosque. Then once you get to know them, and you feel curious to know more, you get close, and they feel that. As soon as they feel that you are so close, they start to invite you to their different activities: meetings, occasions, parties, and so on.”6

The gradual developing of such bonds has likewise been highlighted by other scholars of Islamism. Wickham(2002) describes a similar process in her research, noting that “Islamic ideological outreach was typically a personal, even intimate process, rooted in face-to-face

6Interview, Cairo, August 22, 2012.

7 human relationships. [...] By far the most important institutional vehicle of outreach was the independent mosque” (p. 132-4). Notably, these individual bonds are replicating in mosques across the country, producing personal, local bonds on a national scale. These social bonds, in turn, can be leveraged and mobilized for elections. Decades of research into voter turnout in the United States has pinpointed the enormous importance of social networks and personal appeals from friends and neighbors in mobilizing voters to the polls (i.e., Rolfe, 2012; Green and Gerber, 2019). Mosques, much like churches in the United States (Liu, Austin and Orey, 2009; Gerber, Gruber and Hungerman, 2015), can help to build these social networks and create the ties that later can be activated for elections (Jamal, 2005; Moutselos, 2020). In short, even when Islamists are not preaching from the top down, mosques may still help Islamists develop these horizontal ties. While I am not denying that other mechanisms such as preaching may operate in certain contexts, these personal interactions and social bonds provide a more universal mechanism linking mosques to Islamist votes. The country’s preexisting religious infrastructure thus helps to pave the way for Islamists’ electoral victories. Importantly, secular parties lack a comparable infrastructure. These is no preexisting, nationwide network of institutions that regularly brings together the less religious segment of society, in the way that the mosque does for the more religious. Mosques thus provide a unique infrastructure advantage for Islamists. While intuitive, few studies have attempted to measure and test Islamists’ infrastructure advantage.7 Masoud(2014, p. 169-174) takes a first stab, correlating the number of mosques per capita at the qism level in Cairo and Alexandria with Islamist vote share, but finds only

7One recent study outside of the Middle East finds that waqf -endowed religious institutions, including mosques and religious schools, contributed to the rise of Islamists in Indonesia (Bazzi, Koehler-Derrick and Marx, 2020). Meanwhile, outside of Islam, Freedman(2020) finds that Jewish religious institutions increase support for ultra-Orthodox parties in Israel.

8 mixed evidence.8 He also finds no evidence that self-reported mosque attendance in a 2014 survey correlates with Islamist votes (p. 177). He does however find evidence in Tunisia of a correlation between mosques and Ennahda’s 2011 victory, but underscores that he cannot control for underlying levels of religiosity (p. 196-99). In this paper, I provide two additional tests of the infrastructure advantage. First, I leverage an original dataset of all mosques in Tunisia to provide an ecological correlation between mosque density and Islamist vote share across multiple Tunisian elections. While ecological, I supplement this analysis with original survey and satellite data to rule out alternative explanations, particularly regarding underlying religiosity, politicized imams, and social service provision. Second, I examine region-wide survey data from the Arab Barometer to show that self-reported mosque attendance is indeed a major predictor of trust in Islamists. This individual-level data provides a fine-grained yet region-wide test of our theory, and allows us to control for a number of counter-explanations. While each of these approaches may have their limitations, the consistent results across these two tests lend important empirical evidence that Islamists may indeed benefit from an infrastructure advantage.

4 Mosques in Tunisia

The first method of testing whether mosques give Islamists an advantage in elections is through an ecological analysis of mosque density and Islamist vote share in Tunisia. There are several reasons why Tunisia is a helpful case for testing our theory in this way. First, Tunisia offers multiple waves of free and fair parliamentary elections, including in 2011, 2014, and 2019. In October 2011, in the aftermath of the Arab Spring revolution, the Islamist party Ennahda (the Renaissance) won the country’s first free and fair elections with 37% of the vote. That victory came as a surprise, both because Tunisia had long been

8In a working paper, Brooke and Masoud(2016) revisit the link between mosques and Islamist vote share in Egypt’s 2012 election.

9 viewed as one of the most secular countries in the Arab world, and also because Ennahda as an organization had been largely decimated by former President , leaving them in exile, in prison, or underground (Wolf, 2017; McCarthy, 2018; Grewal, 2020). Then, after governing for three years in the government, Ennahda placed second in the 2014 elections with 27.8% of the vote. Finally, in 2019, Ennahda once again placed first, but with a continually declining vote share, 19.6%. In addition, the 2019 elections also saw the rise, for the first time, of additional Islamist parties that also won seats in parliament. The (‘itila¯f al-kara¯ma), a pro- revolution, more hardline Islamist party led by Seifeddine Makhlouf, a lawyer who defended members of Ansar al-Sharia in court, placed 5th with 5.94% of the vote and 21 seats in parliament. Meanwhile, the salafist Mercy Party (Hizb ) ran in just a handful of districts, winning 1.4% of the vote and 4 parliamentary seats. Tunisia thus provides us with a considerable amount of data to test the robustness of our theory across multiple elections and multiple Islamist parties. Second, the multiple waves allow us to examine the relative importance of the infras- tructure hypothesis compared to other explanations over time. A number of scholars have argued that Islamists should see a declining vote share after the first elections (i.e., Cam- mett and Luong, 2014, p. 202). While Islamists can ride to victory initially based on their clean reputations and promises, after their first victory, they will be judged on their track records. Given their inexperience in governance, these track records might undermine their reputation for competence, while voters may also realize that Islamists are not redistributing wealth. Over time, their compromises may also tarnish their ideological and revolutionary sheen. Indeed, Ennahda’s declining vote share and popularity over the three consecutive parliamentary (Figure2) seems to affirm these views. However, one constant across multiple elections is the infrastructure hypothesis, which should continue to aid voter outreach and mobilization even after the Islamists’ reputation has been tarnished. We therefore might expect the infrastructure advantage to become

10 Figure 2: Ennahda’s Declining Performance

relatively more important over time, as other advantages decline. Third, Tunisia is also a useful case for helping to isolate why there is a mosque effect. While existing literature tends to focus on Islamist imams preaching in the mosque, the case of Tunisia helps to rule out this alternative mechanism, particularly by the 2019 elections. This charge was repeatedly leveled by secularists against Ennahda after its victory in the 2011 elections. But with the change in government after the 2014 elections, and with the 2014 constitution empowering the state to ensure “the neutrality of mosques and places of worship away from partisan instrumentalization” (article 6), Tunisia has cracked down on politicized imams, monitoring and firing imams ahead of the 2018 municipal and 2019 parliamentary elections (Gall, 2015; Asharq al-Awsat, 2018; Donker, 2019; Saidani, 2019). Moreover, Ennahda in May 2016 decided to formally specialize into a , excising any preaching or proselytizing roles from the movement (Marks, 2016). In particular, it internally voted to expel any leader from the party if they also serve as an imam or preach in a mosque, as a signal that it has separated religion from politics (Merone, 2019; Meddeb, 2019). The 2019 elections, occurring after this formal split, should thus help to exclude the possibility of any mosque-Islamist relationship being confounded by politicized imams – though we will make sure to test the proposition nonetheless. Relatedly, the case of Tunisia also helps to address another possible mechanism link- ing mosques to Islamist votes: social service provision. Some Islamists, such as Egypt’s

11 Muslim Brotherhood and Lebanon’s , engage extensively in social service provi- sion, often in or around mosques (Wickham, 2002; Clark, 2004; Brooke, 2019). Accordingly, any spatial correlation we uncover between mosques and Islamist support in these countries might instead be operating through increased exposure to Islamist social services, not simply face-to-face interactions. In Tunisia, by contrast, Ennahda has not engaged in major social service provision (Brooke, 2019, p. 144). Its organization was almost entirely dismantled by former Presi- dent Ben Ali. Unlike Egypt’s Hosni Mubarak, who half-repressed, half-tolerated the Muslim Brotherhood, Ben Ali’s level of repression did not permit Ennahda to set up medical clinics or welfare organizations (McCarthy, 2018, p. 97). The extent of its social solidarity in the 1990s and 2000s were small-scale, informal collections of charity, typically to support the families of imprisoned Ennahda leaders (Wolf, 2017) or other needy families (Cammett and Luong, 2014, p. 196). By and large, Ennahda had to rebuild its organization from scratch following the 2011 revolution. Even after the revolution, Ennahda has not engaged in any systematic, organizational welfare provision. While local-level charities and clinics have emerged, they are affiliated with particular individuals, rather than the Ennahda party (Sigill`o, 2020; McCarthy, 2018, p. 146). Moreover, as part of its 2016 specialization into a political party, Ennahda also decided to excise these da’wa or charitable activities, expelling any leader from the party if they run a charitable organization. As a result of these developments, the 2019 Tunisian elections, in particular, should have the fewest potential confounders. If there is an effect of mosque density on Islamist vote share in 2019, it will likely be operating through personal, horizontal interactions in the mosque, rather than social services or politicized imams. Tunisia is thus a very useful case both for testing the effect of mosques and for isolating the mechanism through which it operates.

12 Islamist Vote Share

For the outcome variable, I examine the Islamists’ vote share in the 2011, 2014, and 2019 parliamentary elections. Their vote share, or their percent of votes received out of the total number of votes cast, provides a direct measure of our outcome of interest – the electoral performance of Islamists. They thus skirt the potential difficulties that a survey-based attitudinal measure of support might have in translating into actual voting behavior. I collect this data at the delegation level. Administratively, Tunisia is divided first into 24 governorates (wila¯ya¯t) and subsequently into 264 delegations (mu‘tamadiya¯t). Delegations are thus a rough equivalent of US counties, with an average population of about 33,000. Notably, party lists are set at the governorate level, not at the delegation, thus mitigating any potential concern that Ennahda’s perfomance in a delegation might be driven by its fielding of a particularly popular candidate there. All electoral results are obtained from the Tunisian elections commission, the Instance Superiore Independante pour les Elections (ISIE).9 We accordingly have five dependent vari- ables, all at the delegation level: Ennahda’s vote share in 2011, 2014, and 2019; Karama’s vote share in 2019; and the vote share of all Islamists in 2019 (Ennahda, Karama, and Errahma).10 For illustration, Figure3 shows Ennahda’s vote share in the 2011 and 2019 elections. There is considerable variation, ranging in 2019 from a low of 5.7% in () to a high of 65% in (Gabes).11

9Results for 2011 were obtained from Masoud(2014); 2014 from Nate Rosenblatt, and 2019 directly from ISIE’s website. 10I do not create a separate variable for Errahma’s vote share as it only ran in a small number of delegations. 11As his (adopted) name indicates, Ennahda leader ’s family origins are from Ghannouch, though he himself grew up in neighboring (Gabes).

13 Figure 3: Ennahda Vote Share, 2011 (left) and 2019 (right)

Mosque Density

Our primary independent variable is the number of mosques per capita at the delegation level. Since mosques are continually being built, I collect this variable at two snapshots in time: first in 2011 (from Masoud(2014)) and second in January 2019 (directly from the Ministry of Religious Affairs, Wiza¯rat al-Shu‘u¯n al-D¯iniyya). According to this data, Tunisia had 4716 total mosques in 2011, and 5934 in 2019. In other words, since Ennahda took office in December 2011, over 1000 new mosques have been built or registered. Although Ennahda never headed the Ministry of Religious Affairs, this increase still raises the possibility that Ennahda may have influenced where new mosques were built.12 If Ennahda reallocated mosque construction to its strongholds, it may artificially create a correlation between mosques and Islamist vote share by the 2014 and 2019 elections. Empirically, this does not appear to be the case: there is not a significant correlation between Ennahda’s 2011 vote share and new mosque construction at the delegation level, and it is actually leaning negative, not positive (see appendix, Figure9). However, the possibility

12Ennahda leader Rached Ghannouchi famously told salafis in April 2012 that “Now we have not just a few mosques, we have the Ministry of Religious Affairs.” See Amara(2012).

14 remains, and thus I privilege as the primary independent variable the mosque data from 2011, prior to Ennahda’s time in office. However, I show in the appendix that results are robust to using the 2019 data instead (Table5).

Figure 4: Mosque Density in 2011 (left) and 2019 (right)

For both years, I standardize the number of mosques per delegation by that delegation’s population, as recorded in the most recent census (2014). I thus calculate mosque density as the number of mosques per 10,000 people. Figure4 presents this data in each of Tunisia’s delegations in 2011 and 2019. There is likewise large variation, ranging in 2019 from 1.16 mosques per 10,000 people in Soukra (Ariana), a posh suburb of , to 31.5 mosques per 10,000 people in Matmata (Gabes), a rural town in the south that we will revisit later in this paper.

Covariates

There is a significant bivariate correlation between mosque density and Islamist vote share in all three elections (see appendix, Figure 11), lending credence to the infrastructure advantage. I then test the robustness of this correlation to a variety of covariates. First, a major potential confounder is religiosity: perhaps areas that are more religious build more mosques and vote for Islamists, creating the correlation between them. I address this confounder in two ways.

15 First, I create a survey-based measure of piety by combining six nationally-representative surveys: the four waves of the Arab Barometer in Tunisia (2011, 2013, 2016, and 2018), and two surveys of my own conducted with the same survey firm, One to One for Research and Polling, in 2019 and 2020 (more details below). The only measure of piety in all six surveys is how often respondents pray, a metric that remains quite stable every year: between 60-70% of respondents in each survey said they pray at least once daily (see appendix, Figure 12). Given its stability over time, I combine all six surveys to produce a total of 8203 respondents, or about 31 per delegation, allowing us to make tentative claims about religiosity at the delegation level. As an alternative measure of religiosity, I follow Ciftci, Robbins and Zaytseva(2020) and collect satellite imagery of night-time lights during the month of Ramadan. Since many rituals during Ramadan occur after sunset (including the meals of ifta¯r and suhu¯r, and the taraw¯ih prayers), we should observe an increase in night-time lights in areas with greater religious observance, compared to the months before and after Ramadan. Using publicly available satellite data,13 I thus calculate each delegation’s percent increase in night-time lights during Ramadan compared to the average of the months before and after as a proxy for sub-national devotion (for more details, see appendix). Beyond religiosity, I control for a number of socioeconomic variables thought to shape support for Islamists. Across the region, Islamists tend to perform better in urban, densely populated areas especially among unemployed and lesser educated youth (i.e., Willis, 1997). From the census I thus control for the delegation’s population, percent youth (aged 20- 29), percent of workforce involved in agriculture (as a proxy for urban/rural), the percent unemployed, and the percent with higher education. For the 2019 elections, I am also able to control for a new delegation-level measure of poverty released in 2020 by Tunisia’s National Institute for Statistics.14

13See VIIRS data at: https://payneinstitute.mines.edu/eog/nighttime-lights/. 14See http://www.ins.tn/en/publication/tunisia-poverty-map-septembre-2020.

16 Third, I control for a number of political grievances thought to undergird Ennahda’s support in Tunisia. The abuses of former dictator Zine El Abidine Ben Ali, including police repression and torture, are thought to have fueled support for the opposition, in particular for Ennahda, which bore the brunt of this repression (Nugent, 2020). Relatedly, Ben Ali’s corruption, and his favoritism for his region (the Sahel), were likewise thought to fuel support for the opposition, including Ennahda, whose pious reputation made it seem more pure and less corruptible. To measure these political grievances, I use the number of cases submitted to the Truth and Dignity Commission (Instance V´erit´eet Dignit´e, IVD) in each governorate for a) human rights violations, b) corruption, and c) regional discrimination.15 Fourth, for the 2019 elections, I am also able to control for three alternative mechanisms. This paper contends that mosques facilitate informal contact with potential supporters. However, mosques might also provide a location for Islamists to preach, provide social ser- vices, or explicitly campaign. To rule these out, I draw on three original surveys conducted around the 2019 elections: (1) a face-to-face survey of 1013 Tunisians conducted in Septem- ber 2019, (2) a phone survey of 1200 Tunisians in January 2020, and (3) an online survey of 11,329 Tunisians that ran in between.16 Combining these three surveys produces a total of 13,542 respondents, or about 51 per delegation. The surveys included three questions that help us address each of these alternative expla- nations. First, they asked respondents whether their local imams were instructing mosque-

15Coded from the final IVD report, and standardized by each governorate’s population. If reports that Ennahda used its infrastructure to help victims file cases are true, it might weaken the mosque variable, thus providing a hard test for still finding a mosque effect. 16The first two surveys are the same used to measure piety. All three were conducted jointly with Steven Brooke and Robert Kubinec. The face-to-face and telephone surveys were conducted by One to One for Research and Polling. The online survey recruited respondents through advertisements on Facebook. The surveys are weighted to population demographics. For more details, see appendix.

17 goers to vote in a particular fashion.17 Only 3.8% claimed that their imams did so, in line with the government’s crackdown on politicized imams. Second, the surveys asked whether respondents personally received an offer from any political party of cash, food, or help in getting a government job in the lead-up to the elections.18 If yes, they were then asked which party made that offer. Overall, only 4% claimed to have received such an offer ahead of the elections, and only 1.2% said those offers came from Ennahda (and unsurprisingly, almost all of these accusations came from Ennahda’s opponents). Finally, the surveys asked whether respondents saw any political parties campaigning in or around mosques.19 Overall, 12.5% claimed to see political parties doing so; notably, most of these accusations came from supporters of secular parties, as well. If mosque density remains a significant predictor of Islamist vote share when controlling for these alternative explanations, it would lend greater credence to the theory that the infrastructure advantage might instead run through personal, face-to-face interactions.

Results

Despite these covariates, the relationship between mosque density and Islamist vote share remains statistically significant in all three elections. Figure5 presents the results for 2011 and 2014, and Figure6 for 2019. Across all five dependent variables – Ennahda’s vote share in 2011, 2014, and 2019; Karama’s vote share in 2019; and all Islamists in 2019 – delegations with greater mosque density were consistently more likely to vote Islamist. Substantively, in 2011, every 5 additional mosques per capita translated to about a 2- point higher vote share for Ennahda. In 2014, moreover, 5 additional mosques translated to about a 5-point higher vote share for Ennahda. That the effect size is over twice as large in 2014 is consistent with our expectation that the importance of mosques should increase over

17Reports of politicized imams had no correlation with mosque density (p=0.79). 18Reports of vote-buying had no correlation with mosque density (p=0.54). 19Reports of parties campaigning near mosques decreases with mosque density (p<0.09).

18 time. As Ennahda’s reputation is tarnished by governance and it loses its reputation for competence and redistribution, the infrastructure advantage appears to become relatively more important.

Figure 5: Mosque Density and Ennahda Vote Share, 2011 and 2014

However, in 2019, the effect size drops back down to 2011 levels. This could be the result of this election’s more strict separation of religion, charity, and politics. As mentioned earlier, post-2014 Tunisia had increasingly cracked down on politicized imams, while Ennahda had instituted new internal rules separating politics from preaching and charity. In other words, the significant correlation that remains in 2019 might represent the portion of the effect running solely through personal, horizontal interactions.

19 Figure 6: Mosque Density and Islamist Vote Share in 2019

20 Moreover, the 2019 elections also allow us to explicitly rule out these alternative mech- anisms. Mosque density remains a significant predictor of Islamist vote share in 2019 even when controlling for the (few) reports of Ennahda vote-buying, campaigning near mosques, or politicized imams. None of these variables, meanwhile, ever positively correlate with Is- lamist vote share. While not definitive, these results suggest that the link between mosque density and Islamist vote share might instead be operating through more subtle, personal interactions in the mosque. In addition to mosque density, piety is likewise a consistent predictor of voting for En- nahda, validating our survey-based measure of religiosity.20 Moreover, the fact that mosque density remains significant after controlling for piety suggests that the link between mosques and Islamist vote share is not simply reflecting underlying levels of religiosity. The infrastructure advantage does not appear to be limited to Ennahda. The Karama Coalition likewise sees a boost in its vote share in delegations with greater mosque density (Figure6, middle). Karama, which averaged about 6% of the vote nationally, could expect to receive a full point higher vote share in delegations with 5 additional mosques. Mosque density was likewise a significant predictor of the combined vote share of all the Islamist parties in 2019 (Figure6, bottom). In other words, the infrastructure advantage seems to benefit Islamist parties across the board. Many of the demographic and political covariates are also significant. Across the board, Islamists tended to perform better in urban, more densely populated, and lesser educated areas. The Karama coalition seemed to gain the support of delegations with greater unem- ployment, though curiously not those with greater poverty. In all three elections, Ennahda received significantly more votes from areas with more IVD cases for human rights violations, suggesting that despite largely abandoning transitional justice (Marzouki, 2015; Grewal and Hamid, 2020), Ennahda may continue to receive sympathy for being “co-victims” (Guiler, 2020; Nugent, 2020) of Ben Ali.

20However, the satellite-based proxy is not significant: see appendix, Tables3 and4.

21 Illustrations

To further explore why mosques might be correlating with Islamist votes, we can examine the delegations in which mosques seem to matter the most. In these delegations, where the effect size is largest, the mechanisms should be clearest. Figure7 plots the change in residuals for each delegation when running the regression with and without mosque density. Positive values indicate that the model’s prediction for that delegation’s Islamist vote share improved when taking into account its mosque density. In other words, the delegations in the top-right of the figure are areas where we would not have expected Islamists to do well given their level of religiosity, poverty, etc., but where mosque density helps to explain why they did end up performing well.

Figure 7: Case Selection based on Total Islamist Vote Share in 2019

By this metric, mosques appear to have their greatest effect on Islamist vote share in the delegation Nord. There appears to be at least two reasons why. First, Kairouan Nord is home to the Great Mosque of Kairouan, Tunisia’s oldest, largest, and most famous mosque. The mosque itself is massive: 9000 square meters, considerably larger than a football field. If, as we propose, the importance of a mosque is in gathering together large numbers of pious individuals for Islamists to meet, it stands to reason that this should happen most effectively in one as large as the Great Mosque. Second, even more importantly, Kairouan

22 Nord has by far the most mosques of any delegation in absolute terms – 117 in 2011 and 127 in 2019, 40 more than the next delegation and more than 5 times the average (18 in 2011, 23 in 2019). Mosques should provide Islamists their greatest infrastructure advantage in Kairouan Nord. Some of the other high powered cases tell a similar story. (Medenine), Tameghza (), Sud (Kebili), and Matmata (Gabes), are all small towns in the extreme south of the country. They are very traditional, rural towns, with limited state investment into roads, water, or electricity. Matmata, for instance, is famous for its “sub- terranean” or “troglodyte” homes literally built into the earth. In these towns, the mosque occupies a central place in village life, and is one of the few public spaces for residents to gather. The heightened importance of the mosque in these delegations may explain why mosques produce a relatively larger effect on Islamist vote share here. In short, the data from Tunisia suggest that mosques may lend Islamists an advantage in elections, with mosque density strongly correlating with the vote share of multiple Islamist parties across multiple elections. To see if this effect generalizes beyond Tunisia, I now turn to region-wide data from the Arab Barometer.

5 Arab Barometer

To explore how far the theory travels, I leverage region-wide survey data from the Arab Barometer, which has conducted five waves of nationally representative face-to-face surveys in the Arab world. Three of these waves include all the survey questions we need for our analysis: Wave 2 (2011), Wave 3 (2013), and Wave 5 (2018-19). Wave 1 did not ask about trust in Islamists, while Wave 4 did not ask about mosque attendance. For our dependent variable, we examine respondents’ level of trust in the main Islamist party in their country.21 Wave 2, conducted in 2011, asked this question in Egypt and Tunisia

21The Arab Barometer only asked for vote choice in 2011, and few respondents answered.

23 (N=2415). Wave 3 asked it in all countries surveyed, namely Algeria, Egypt, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Palestine, Sudan, Tunisia, and Yemen (N=14,809). Fi- nally, Wave 5 asked it in Egypt, Iraq, Jordan, Libya, Morocco, Palestine, Sudan, Tunisia, and Yemen (N=10,319). In all countries, trust in Islamists is asked on a four-point scale from absolutely no trust (1) to a great extent of trust (4). We treat this as a continuous variable and run standard OLS regressions, though results are robust to ordered logistic regression. For our primary independent variable, we record how often respondents say they attend Friday prayers (Sala¯t al-Jumu‘ah). Friday is the most holy day in Islam, and the Jumu‘ah prayers the most important of all prayers, equivalent according to one hadith to an entire year of individual worship, prayer, and fasting.22 Notably, unlike other prayers, the Friday prayer must be performed in the mosque; indeed, the more literal translation of Sala¯t al- Jumu‘ah is congregational or communal prayer. Friday prayers are thus the most commonly attended prayer at the mosque, and accordingly a useful measure of mosque attendance.23 Of course, attendance at Friday prayers not only reflects exposure to the mosque, our concept of interest, but also religiosity, which could also increase trust in Islamists. To distinguish between these, we include in every model a control for frequency of prayer in general, which would capture both prayers at the mosque as well as prayers performed individually. Indeed, the correlation between how often one attends Friday prayer and how often one prays in general was just 0.33 in 2011, 0.36 in 2013, and 0.17 in 2019.24 Including a control for frequency of prayer in general thus allows us to more credibly claim that any effect of Friday prayers is operating through mosque attendance, not religiosity. We also control for a number of variables that capture the existing explanations for

22al-Tirmidhi, 496. See also https://seekersguidance.org/answers/hanafi-fiqh/ the-rulings-related-to-friday-prayer/. 23The Friday prayer thus also boasts important mobilization potential, both in the Arab Spring (Ketchley and Barrie, 2020), and in Pakistan (Butt, 2016). 24Even among men, the correlations were just 0.56, 0.63, and 0.39.

24 support for Islamists (for full question wording, see appendix). First, to capture Islamist ideology, we include respondents’ support for shari‘a, antipathy to the US and to Israel, and opposition to the rights of non-Muslims and women. To proxy for a desire for Islamist services and redistribution, we examine respondents’ evaluations of whether their govern- ment is tackling inequality and corruption. To measure Islamist reputations, we measure how highly respondents rank “integrity” and “piety” as qualifications for political leadership (wave 2 and 3), and whether religious leaders are less corrupt than non-religious ones (wave 5). For a reputation for reforming the system, we include respondents’ support for democ- racy. Finally, we include standard demographic variables including age, gender, marriage, education, income, unemployment, and country fixed effects. Figure8 presents the results for Wave 2 (top), Wave 3 (middle), and Wave 5 (bottom) of the Arab Barometer. Across all three waves, attending the mosque for Friday prayers is a major predictor of trust in Islamists. In 2011, for instance, an individual who “always” attends Friday prayers was about a half a point more trusting of Islamists (on the 1-4 point scale) than someone who “never” attends Friday prayer. Notably, this effect is highly significant (p<0.001) despite controlling for frequency of prayer, suggesting that mosque attendance is having an additional, independent effect from religiosity. Moreover, mosque attendance is one of the few variables that is consistently significant across all three waves of the Arab Barometer. The others are support for sharia, the hallmark of the Islamist agenda, and support for democracy.25 The importance of mosque attendance does not appear to be limited to men: an interaction by gender is only (weakly) significant in one of the three waves (see appendix, Table7). In short, mosque attendance, as proxied by Friday prayers, appears to be one of the most important predictors of trust in Islamists.

25Corruption is consistently in the opposite direction, with respondents thinking there is corruption in the government being less supportive of Islamists.

25 Figure 8: Friday Prayers at the Mosque and Trust in Islamists (Arab Barometer)

26 6 Conclusion

This paper sought to test whether Islamists might win elections in the Middle East in part due to an ‘infrastructure advantage.’ Islamists, we argue, might be able to leverage the preexisting, nationwide network of mosques to meet new supporters and deepen connections through repeated interactions. Because mosques gather together relatively pious individuals, this infrastructure advantage should uniquely benefit Islamists moreso than secularists. This paper tested this theory in two ways. First, it examined delegation-level data in Tunisia, showing that the number of mosques per capita strongly correlated with the vote share of multiple Islamist parties and across multiple elections. While ecological, it drew upon a wide variety of census, survey, and satellite data to rule out potential confounders and alternative mechanisms. Second, it showed through region-wide survey data that this relationship might generalize beyond Tunisia, with mosque-goers tending to be more trusting of Islamists across the Middle East, even when controlling for religiosity and a number of other sources of the Islamist appeal. The paper thus provides an important corrective or qualification to recent literature that highlights the appeal of Islamists, whether their ideologies, services, or reputations. Instead, this paper revives an earlier literature by social movement theorists that emphasizes that Islamists may also be better resourced than other parties. Islamists appear to be able to ride the preexisting religious infrastructure to electoral victory. While beyond the scope of this article, the findings in this paper likely generalize beyond the Middle East and beyond Islamist parties. Bazzi, Koehler-Derrick and Marx(2020) find that religious institutions correlate with Islamist vote share in Indonesia, while Freedman (2020) uncovers a similar effect in Israel for ultra-Orthodox parties. It stands to reason that religious infrastructure might likewise give religious parties or movements an advantage elsewhere, such as for Hindu nationalists in India or the religious right in the United States. This paper likewise carries important implications for governments in the Middle East. As previously mentioned, dictators across the region have attempted to stem the Islamists’

27 rise by regulating imams and their sermons, and by shuttering mosques in between prayers. Yet while these moves may repress the most explicit forms of Islamist outreach, our results suggest that they only partially mitigate the Islamist advantage. Finally, the results in this paper may also inspire or motivate a future research agenda on the politics of mosque construction. Given the important advantage mosques lend to Islamists, it is possible that governments, political parties, donors, and foreign powers might be cognizant of these potential downstream consequences when deciding where to build mosques. In Tunisia, for instance, supplementary analyses (see appendix, figures9 and 10) suggest that new mosques built after Ennahda came to power in 2011 were disproportion- ately built in areas where Ennahda was weak in 2011, and that these new mosques might have boosted Ennahda’s electoral fortunes in later elections. While speculative, it begs the question of whether Ennahda leaders might have influenced mosque construction (hence jus- tifying our decision to use the 2011 mosque data rather than 2019 for our main analyses). But moving forward, these initial results call attention to the important yet understudied politics that might underlie mosque construction.

28 References

Abu-Amr, Ziad. 1993. “Hamas: A Historical and Political Background.” Journal of Palestine Studies 22(4):5–19. al Anani, Khalil. 2020. “Devout Neoliberalism?! Explaining Egypt’s Muslim Brotherhood’s Socioeconomic Perspective and Policies.” Politics and Religion Firstview.

Amara, Tarek. 2012. “Tunisia islamist leader stirs fears of radicalism in video.” Reuters October 11.

URL: https: // www. reuters. com/ article/ uk-tunisia-ghannouchi-salafis- idUKBRE89A16G20121011? edition-redirect= uk

Asharq al-Awsat. 2018. “Tunis ta‘ud khita li-‘tahiid’ al-masajid khilal intikhabat 2019 [Tunisia prepares plan to ‘neutralize’ mosques during the 2019 elections].” December 17. URL: https://bit.ly/3oepgD4

Ayubi, Nazih. 1991. Political Islam: Religion and Politics in the Arab World. Routledge.

Bazzi, Samuel, Gabriel Koehler-Derrick and Benjamin Marx. 2020. “The Institutional Foun- dations of Religious Politics: Evidence from Indonesia.” Quarterly Journal of Economics 135(2):845–911.

Ben Salem, Maryam. 2020. “‘God loves the rich.’ The Economic Policy of Ennahda: Liber- alism in the Service of Social Solidarity.” Politics and Religion Firstview.

Boulby, Marion. 1988. “The Islamic Challenge: Tunisia since Independence.” Third World Quarterly 10(2):590–614.

Brooke, Steven. 2017. “From Medicine to Mobilization: Social Service Provision and the Islamist Reputational Advantage.” Perspectives on Politics 15(1):42–61.

Brooke, Steven. 2019. Winning Hearts and Votes: Social Services and the Islamist Political Advantage. Cornell University Press.

29 Brooke, Steven and Tarek Masoud. 2016. “Face-to-Face in the House of God: Religious Infrastructure and Voter Mobilization in Egypt’s 2012 Presidential Elections.” Working Paper .

Butt, Ahsan I. 2016. “Street Power: Friday Prayers, Islamist Protests, and Islamization in Pakistan.” Politics and Religion 9:1–28.

Cammett, Melani. 2014. Compassionate Communalism: Welfare and Sectarianism in Lebanon. Cornell University Press.

Cammett, Melani and Pauline Jones Luong. 2014. “Is There an Islamist Political Advan- tage?” Annual Review of Political Science 17:187–206.

Cammett, Melani and Sukriti Issar. 2010. “Bricks and Mortar Clientelism Sectarianism and the Logics of Welfare Allocation in Lebanon.” World Politics 62(3):381–421.

Ciftci, Sabri, Michael Robbins and Sofya Zaytseva. 2020. “Devotion at Sub-National Level: Ramadan, Nighttime Lights, and Religiosity in the Egyptian Governorates.” International Journal of Public Opinion Research Forthcoming:1–19.

Clark, Janine. 2004. Islam, Charity, and Activism:Middle-Class Networks and Social Welfare in Egypt, Jordan, and Yemen. Indiana University Press.

Donker, Teije H. 2019. “The Sacred as Secular: State Control and Mosques Neutrality in Post-Revolutionary Tunisia.” Politics and Religion 12.

Esposito, John. 1997. Political Islam: Revolution, Radicalism or Reform. Lynne Rienner.

Freedman, Michael. 2020. “Vote with Your Rabbi: The Electoral Effects of Religious Insti- tutions in Israel.” Electoral Studies Forthcoming:1–33.

Gall, Carlotta. 2015. “Tunisia’s Secular Government Cracks Down on Mosques in Aftermath of Massacre.” New York Times July 23.

30 Gerber, Alan S., Jonathan Gruber and Daniel M. Hungerman. 2015. “Does Church Atten- dance Cause People to Vote? Using Blue Laws’ Repeal to Estimate the Effect of Religiosity on Voter Turnout.” British Journal of Political Science 46:481–500.

Gidengil, Elisabeth and Ekrem Karakoc. 2016. “Which matters more in the electoral success of Islamist (sucessor) parties - religion or performance? The Turkish case.” Party Politics 22(3):325–338.

Green, Donald P. and Alan S. Gerber. 2019. Get Out the Vote: How to Increase Voter Turnout. Brookings Institution Press.

Grewal, Sharan. 2020. “From Islamists to Muslim Democrats: The Case of Tunisia’s En- nahda.” American Political Science Review 114(2):519–535.

Grewal, Sharan, Amaney A. Jamal, Tarek Masoud and Elizabeth R. Nugent. 2019. “Poverty and Divine Rewards: The Electoral Advantage of Islamist Political Parties.” American Journal of Political Science 63(4):859–874.

Grewal, Sharan and Shadi Hamid. 2020. “The Dark Side of Consensus in Tunisia: Lessons from 2015-2019.” Brookings Institution . URL: https://www.brookings.edu/research/the-dark-side-of-consensus-in-tunisia-lessons- from-2015-2019/

Guiler, Kim. 2020. “From prison to parliament: Victimhood, identity, and electoral support.” Mediterranean Politics Firstview:1–30.

Haddad, Yvonne. 1992. “Islamists and the ‘Problem of Israel’: The 1967 Awakening.” Middle East Journal 46(2):266–285.

Jamal, Amaney. 2005. “The Political Participation and Engagement of Muslim Americans: Mosque Involvement and Group Consciousness.” American Politics Research 33(4):521– 544.

31 Kandil, Hazem. 2015. Inside the Brotherhood. Polity Press.

Kepel, Gilles. 2002. Jihad: The Trail of Political Islam. I.B. Tauris.

Ketchley, Neil and Christopher Barrie. 2020. “Fridays of Revolution: Focal Days and Mass Protest in Egypt and Tunisia.” Political Research Quarterly 73(2):308–324.

Liu, Baodong, Sharon D. Wright Austin and Byron D’Andra Orey. 2009. “Church At- tendance, Social Capital, and Black Voting Participation.” Social Science Quarterly 90(3):576–592.

Marks, Monica. 2016. “How big were the changes Tunisia’s Ennahda party just made at its national congress?” Washington Post May 25. URL: https://www.washingtonpost.com/news/monkey-cage/wp/2016/05/25/how-big- were-the-changes-made-at--ennahda-just-made-at-its-national-congress/

Marzouki, Nadia. 2015. “Tunisia’s Rotten Compromise.” Middle East Research and Infor- mation Project . URL: http://www.merip.org/mero/mero071015

Masoud, Tarek. 2014. Counting Islam: Religion, Class, and Elections in Egypt. Cambridge University Press.

McCarthy, Rory. 2018. Inside Tunisia’s Al-Nahda: Between Politics and Preaching. Cam- bridge University Press.

Meddeb, Hamza. 2019. “Ennahda’s Uneasy Exit from Political Islam.” Carnegie Middle East Center September 5. URL: https://carnegie-mec.org/2019/09/05/ennahda-s-uneasy-exit-from-political-islam- pub-79789

Merone, Fabio. 2019. “Politicians or Preachers? What Ennahda’s Transformation Means for Tunisia.” Carnegie Middle East Center January 31.

32 URL: https://carnegie-mec.org/2019/01/31/politicians-or-preachers-what-ennahda-s- transformation-means-for-tunisia-pub-78253

Moutselos, Michalis. 2020. “Praying on Friday, voting on Sunday? Mosque attendance and voter turnout in three West European democracies.” Journal of Ethnic and Migration Studies 47(11):2275–2292.

Nugent, Elizabeth R. 2020. After Repression: How Polarization Derails Democratic Transi- tion. Princeton University Press.

Nugent, Elizabeth R., Tarek Masoud and Amaney A. Jamal. 2018. “Arab Responses to Western Hegemony: Experimental Evidence from Egypt.” Journal of Conflict Resolution 62(2):254–288.

Pellicer, Miquel and Eva Wegner. 2014. “Socio-economic voter profile and motives for Islamist support in Morocco.” Party Politics 20(1):116–133.

Pepinsky, Thomas B., R. William Liddle and Saiful Mujani. 2012. “Testing Islam’s Political Advantage: Evidence from Indonesia.” American Journal of Political Science 56(3):584– 600.

Pew Research Center. 2013. “The World’s Muslims: Religion, Politics and Society.” The Pew Forum on Religion & Public Life April 30. URL: https://www.pewforum.org/2013/04/30/the-worlds-muslims-religion-politics- society-overview/

Rolfe, Meredith. 2012. Voter Turnout: A Social Theory of Political Participation. Cambridge University Press.

Saidani, Mongi. 2019. “Campaign to Steer Tunisia Mosques Away from Political Cam- paigns.” Asharq al-Awsat April 20.

33 URL: https://english.aawsat.com//home/article/1687451/campaign-steer-tunisia- mosques-away-political-campaigns

Shehata, Samer and Joshua Stacher. 2006. “The Brotherhood goes to Parliament.” Middle East Research and Information Project 240.

Sigill`o,Ester. 2020. “Islamism and the rise of Islamic charities in post-revolutionary Tunisia: claiming political Islam through other means?” British Journal of Middle Eastern Studies Forthcoming.

Tessler, Mark. 2010. “Religion, Religiosity, and the Place of Islam in Political Life: Insights from the Arab Barometer Surveys.” Middle East Law and Governance 2:221–252.

Trager, Eric. 2016. Arab Fall: How the Muslim Brotherhood Won and Lost Egypt in 891 Days. Georgetown University Press.

Waltz, Susan. 1986. “Islamist Appeal in Tunisia.” Middle East Journal 40(4):651–670.

Wedeen, Lisa. 2003. “Beyond the Crusades: Why Huntington, and Bin Laden, Are Wrong.” Middle East Policy 10(2):54–61.

Wickham, Carrie Rosefsky. 2002. Mobilizing Islam: Religion, Activism and Political Change in Egypt. Columbia University Press.

Wiktorowicz, Quintan. 1998. “State Power and the Regulation of Islam in Jordan.” Journal of Church and State 41(4):677–696.

Wiktorowicz, Quintan. 2004. Islamic Activism: A Social Movement Theory Approach. Indi- ana University Press.

Willis, Michael. 1997. The Islamist Challenge in Algeria: A Political History. New York University Press.

34 Wolf, Anne. 2017. Political : The History of Ennahda. Oxford University Press.

Yildirim, A. Kadir and Caroline M. Lancaster. 2015. “Bending with the Wind: Revisiting Islamist Parties’ Electoral Dilemma.” Politics and Religion 8:588–613.

35 7 Appendix 1: Mosques in Tunisia

Mosque construction, 2011-2019

The 1000 new mosques built or registered between 2011 and 2019 were not disproportionately allocated to Ennahda strongholds. Figure9 plots the percent increase in mosques between 2011 and 2019 in each governorate (top) and delegation (bottom) according to Ennahda’s vote share in 2011. If Ennahda rewarded its strongholds, we should see a positive relationship, with more mosques being built in areas where Ennahda did well in 2011.

Figure 9: Strategic construction of mosques, 2011 to 2019

On the contrary, we find a statistically significant negative correlation, particularly visible at the governorate level. New mosques after 2011 were built in areas where Ennahda was

36 weak, not strong. This correlation holds when controlling for population growth during this time period, suggesting this skewed increase in mosques was not driven by ‘objective’ criteria such as overcrowding in existing mosques. It is thus possible, though speculative, that Ennahda leaders might have known that mosques help them in elections, and so strategically allocated new mosques to areas where they were weak. Indeed, further analyses suggest that these new mosques built after 2011 might have boosted Ennahda’s electoral fortunes in 2019. Generally, Ennahda’s support declined in all governorates between 2011 and 2019. But, figure 10 shows that governorates that saw large increases in mosque density between 2011 and 2018 tended to see more stable vote shares for Ennahda. While this is simply a bivariate relationship, it raises a potential future research agenda exploring the politics of mosque construction.

Figure 10: New mosques and Ennahda vote share, 2011 to 2019

Bivariate relationships

Figure 11 plots the bivariate correlations between 2011 mosque density and the total Islamist vote share in the 2011, 2014, and 2019 elections. As can be seen, there is a positive and significant association in all three elections. Islamists performed significantly better in areas with higher mosque density.

37 Figure 11: Bivariate Relationship, Mosques and Islamist Vote Share

Notably, the effect of mosque density grows considerably stronger from 2011 to 2014, in line with our hypothesis that it should increase in importance over time as Islamists’ reputations are tarnished by governance. Substantively, in 2014, every 1 additional mosque per 10,000 residents translated into a full point higher vote share for Ennahda, twice as large as it did in 2011. However, the effect of mosque density, while still significant, appears to weaken by 2019, falling to every 1 additional mosque translating to about half a point higher vote share. This could be the result of the state’s post-2014 crackdown on politicized imams as well as Ennahda’s internal moves to separate politics from preaching and charity. If so, the significant correlation that remains might represent the portion of the effect running instead through personal, horizontal interactions.

38 Piety (Survey)

Table1 shows the dates and sample sizes for the six nationally-representative surveys I combine to generate a delegation-level measure of piety. All surveys were conducted by the same Tunisian firm, One to One for Research & Polling.

Table 1: Surveys to measure piety (N=8203)

Survey Year Dates Sample Mode Arab Barometer Wave 2 2011 Sep 30 – Oct 11 1196 Face-to-Face Arab Barometer Wave 3 2013 Feb 3 – Feb 25 1199 Face-to-Face Arab Barometer Wave 4 2016 Feb 13 – Mar 3 1200 Face-to-Face Arab Barometer Wave 5 2018 Oct 29 – Dec 4 2400 Face-to-Face Original 2019 Sep 9 – Sep 14 1008 Face-to-Face Original 2020 Jan 11 – Jan 24 1200 Telephone

The survey-based measure of piety (percent of respondents who pray daily) is relatively stable across all six surveys (Figure 12), and hence we use this aggregate measure for all three elections.

Figure 12: Survey-based measure of piety, 2011 to 2020

39 Piety (Satellite)

Night-time lights data are publicly available on a monthly basis from 2012.26 Ramadan does not always align with the Gregorian calendar, but fortuitously did so in both 2014 (June 28-July 28) and 2019 (May 5-June 3). In 2014, I thus compare night-time lights in July to June and August, and for 2019, May compared to April and June. Given that satellite data are not available for 2011, I use the 2014 data for both the 2011 and 2014 elections, and the 2019 data for the 2019 elections. We first clip the (already) stray-light corrected data to Tunisian delegations, and then, following Ciftci, Robbins and Zaytseva(2020), pre-process the data further by 1) removing negative values (noise), 2) removing extreme values (anything above the maximum in Tunis, which ends up removing what appear to be oil refineries in and Medenine), and then 3) imputing those missing values with the average of the 8 surrounding pixels. For each month, we then compute each delegation’s total radiance (the sum of all pixel values) divided by the delegation’s square kilometers. We compare this average radiance per km2 in Ramadan to the average of the months before and after to create our proxy for sub-national religiosity. Results are similar when comparing just to the month before, or just to after.

Mechanisms

To explore the mechanisms, we leverage three surveys conducted in Tunisia before and after the 2019 elections. Two are the September 2019 (N=1008) and January 2020 (N=1200) surveys listed above, while the third is an online survey of 11,329 Tunisians that ran in between and recruited respondents through advertisements on Facebook. The advertisement (Figure 13) invited Tunisians to “take an academic survey about the Tunisian elections.” Given that only 62% of Tunisians were on Facebook at the time of the online survey,27

26See https://eogdata.mines.edu/nighttime_light/monthly/v10/. 27See https://napoleoncat.com/stats/facebook-users-in-tunisia/2020/01.

40 Figure 13: Facebook advertisment it is not representative. As Table2 shows, it skews more male, younger, better educated, and wealthier. We accordingly weight the surveys to bring them in line with the population. Table2, final column, presents the weighted data.

Table 2: Demographics and Mechanisms from Tunisia Surveys (%)

Face-to-Face Phone Facebook Total Weighted (N=1008) (N=1200) (N=11,329) (N=13,542) (N=13,542) Demographics Female 48.5 50.0 41.3 42.6 50.0 Age < 30 22.6 20.7 33.9 31.9 21.0 Unemployed 17.0 14.6 15.4 15.4 16.0 Univ. graduate 14.6 16.5 60.6 53.2 15.0 > 2000 TND/month 5.1 4.7 24.3 21.1 5.0 Mechanisms Politicized Imam 2.3 2.3 5.1 4.5 3.8 Vote-buying (Islamist) 0.4 21.7 1.2 1.3 1.2 Campaigning 20.6 16.3 10.4 12.2 12.5

We use the following survey questions to test alternative mechanisms:

• Did your local imam encourage you to vote for a particular candidate?

41 • Which of the following kinds of contact have you received from Tunisian political parties and/or presidential candidates? Please check all that apply.

– Offer of cash in exchange for support

– Offer of gifts of foods or other necessities

– Offer of help with local bureaucracy/obtaining jobs

• [If yes to any of the above] From which parties did you receive these offers?

• Were any political candidates campaigning inside or around your local mosque?

Regression Tables

Tables3 and4 are the regressions from which Figures5 and6 were created. Table5 shows that the 2019 results are robust to using the 2019 mosque data.

42 Table 3: Mosque Density and Ennahda’s Vote Share, 2011 and 2014 (OLS)

Dependent variable: Ennahda Vote Share 2011 2014 (1) (2) (3) (4) Mosques per 10,000 0.454∗∗ 0.505∗∗∗ 0.930∗∗∗ 0.940∗∗∗ (0.180) (0.174) (0.216) (0.213)

Piety (Survey) 9.838∗∗ 18.639∗∗∗ (4.371) (5.287)

Piety (Satellite) 6.930 4.967 (8.773) (11.334)

Population 0.744∗∗ 0.802∗∗∗ 0.845∗∗ 0.758∗∗ (0.304) (0.304) (0.371) (0.383)

% age 20-29 −0.254 −0.361 0.132 −0.134 (0.252) (0.251) (0.336) (0.352)

% Agriculture −0.298∗∗∗ −0.312∗∗∗ −0.216∗∗∗ −0.289∗∗∗ (0.039) (0.039) (0.070) (0.072)

Unemployment −0.223∗∗ −0.274∗∗ 0.160 0.030 (0.110) (0.106) (0.146) (0.146)

% Higher Edu −0.473∗∗∗ −0.550∗∗∗ −0.353∗∗ −0.517∗∗∗ (0.122) (0.121) (0.141) (0.148)

IVD Human Rights 0.726∗∗∗ 0.710∗∗∗ 1.029∗∗∗ 1.035∗∗∗ (0.172) (0.166) (0.209) (0.212)

IVD Corruption 0.086 0.104∗∗ −0.082 −0.035 (0.056) (0.052) (0.094) (0.083)

IVD Regional −0.099 −0.055 0.322 0.641 (0.293) (0.292) (0.439) (0.447)

Constant 36.243∗∗∗ 46.003∗∗∗ 1.943 23.060∗∗∗ (6.405) (5.503) (8.028) (7.431)

Observations 240 248 214 223 R2 0.381 0.382 0.383 0.342 Adjusted R2 0.354 0.356 0.352 0.311 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

43 Table 4: Mosque Density and Islamist Vote Share in 2019 (OLS)

Dependent variable: 2019 Vote Share Ennahda Karama All Islamists (1) (2) (3) (4) (5) (6) Mosques per 10,000 0.421∗∗∗ 0.449∗∗∗ 0.238∗∗∗ 0.200∗∗∗ 0.584∗∗∗ 0.586∗∗∗ (0.159) (0.159) (0.076) (0.073) (0.199) (0.196)

Piety (Survey) 11.149∗∗∗ −1.974 9.642∗ (3.939) (1.887) (4.930)

Piety (Satellite) −3.679 −8.952 −10.502 (16.174) (7.396) (19.977)

Population 0.267 0.205 0.364∗∗∗ 0.334∗∗∗ 0.769∗∗∗ 0.689∗∗ (0.230) (0.240) (0.110) (0.110) (0.289) (0.296)

% age 20-29 0.259 0.064 0.150 0.105 0.449 0.209 (0.230) (0.241) (0.110) (0.110) (0.288) (0.298)

% Agriculture −0.142∗∗ −0.184∗∗∗ 0.015 0.017 −0.118 −0.161∗∗ (0.057) (0.060) (0.027) (0.027) (0.072) (0.074)

Unemployment −0.120 −0.222∗ 0.249∗∗∗ 0.210∗∗∗ 0.193 0.044 (0.117) (0.116) (0.056) (0.053) (0.146) (0.143)

Poverty −0.023 −0.045 −0.193∗∗∗ −0.172∗∗∗ −0.253∗∗∗ −0.253∗∗∗ (0.073) (0.076) (0.035) (0.035) (0.092) (0.094)

% Higher Edu −0.345∗∗∗ −0.458∗∗∗ −0.125∗∗ −0.100∗ −0.469∗∗∗ −0.570∗∗∗ (0.113) (0.118) (0.054) (0.054) (0.141) (0.146)

IVD Human Rights 0.777∗∗∗ 0.799∗∗∗ 0.029 0.037 0.691∗∗∗ 0.744∗∗∗ (0.158) (0.161) (0.075) (0.074) (0.197) (0.199)

IVD Corruption −0.092∗ −0.050 −0.094∗∗∗ −0.081∗∗∗ −0.198∗∗∗ −0.138∗∗ (0.051) (0.049) (0.024) (0.022) (0.064) (0.061)

IVD Regional 0.575∗ 0.791∗∗ −0.436∗∗∗ −0.339∗∗ −0.064 0.255 (0.296) (0.305) (0.142) (0.140) (0.371) (0.377)

Political Imam 9.560 6.516 0.361 −2.467 10.531 4.337 (7.555) (7.212) (3.619) (3.298) (9.456) (8.907)

Vote-Buying −27.619∗∗ −26.440∗∗ 2.733 3.248 −28.741∗ −26.611 (12.796) (13.335) (6.131) (6.098) (16.017) (16.470)

Campaigning 7.343 3.524 0.660 0.238 10.089 5.272 (4.910) (4.534) (2.352) (2.073) (6.145) (5.600)

Constant 5.347 19.824∗∗∗ 2.639 2.652 8.153 22.893∗∗∗ (5.848) (5.322) (2.802) (2.434) (7.320) (6.573)

Observations 238 246 238 246 238 246 R2 0.320 0.289 0.284 0.270 0.295 0.269 Adjusted R2 0.278 0.246 0.239 0.226 0.251 0.224 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

44 Table 5: Using 2019 mosque density (OLS)

Dependent variable: 2019 Vote Share Ennahda Karama All Islamists (1) (2) (3) (4) (5) (6) Mosques per 10,000 0.298∗ 0.376∗∗ 0.175∗∗ 0.161∗∗ 0.399∗∗ 0.474∗∗ (0.152) (0.153) (0.073) (0.070) (0.191) (0.190)

Piety (Survey) 11.339∗∗∗ −1.961 9.888∗∗ (3.936) (1.896) (4.948)

Piety (Satellite) −5.361 −9.437 −12.259 (16.252) (7.459) (20.141)

Population 0.191 0.155 0.336∗∗∗ 0.318∗∗∗ 0.678∗∗ 0.632∗∗ (0.232) (0.242) (0.112) (0.111) (0.292) (0.300)

% age 20-29 0.363 0.177 0.186 0.139 0.558∗ 0.324 (0.236) (0.249) (0.114) (0.114) (0.297) (0.309)

% Agriculture −0.139∗∗ −0.177∗∗∗ 0.016 0.019 −0.115 −0.153∗∗ (0.057) (0.060) (0.027) (0.027) (0.071) (0.074)

Unemployment −0.119 −0.210∗ 0.245∗∗∗ 0.210∗∗∗ 0.185 0.050 (0.117) (0.117) (0.056) (0.054) (0.147) (0.145)

Poverty −0.034 −0.058 −0.200∗∗∗ −0.179∗∗∗ −0.270∗∗∗ −0.272∗∗∗ (0.073) (0.076) (0.035) (0.035) (0.092) (0.094)

% Higher Edu −0.339∗∗∗ −0.440∗∗∗ −0.125∗∗ −0.096∗ −0.469∗∗∗ −0.555∗∗∗ (0.114) (0.120) (0.055) (0.055) (0.143) (0.149)

2011 Rev Martyrs −1.526∗ −1.513∗ −0.488 −0.413 −1.588 −1.495 (0.803) (0.839) (0.387) (0.385) (1.010) (1.040)

IVD Human Rights 0.852∗∗∗ 0.865∗∗∗ 0.056 0.059 0.780∗∗∗ 0.817∗∗∗ (0.159) (0.163) (0.077) (0.075) (0.200) (0.202)

IVD Corruption −0.085∗ −0.046 −0.091∗∗∗ −0.079∗∗∗ −0.188∗∗∗ −0.132∗∗ (0.051) (0.049) (0.025) (0.023) (0.064) (0.061)

IVD Regional 0.541∗ 0.739∗∗ −0.445∗∗∗ −0.353∗∗ −0.091 0.205 (0.300) (0.309) (0.144) (0.142) (0.377) (0.383)

Political Imam 9.761 6.738 0.556 −2.301 10.861 4.744 (7.578) (7.236) (3.651) (3.321) (9.527) (8.967)

Vote-Buying −29.507∗∗ −28.516∗∗ 1.889 2.483 −30.979∗ −29.011∗ (12.844) (13.389) (6.188) (6.145) (16.147) (16.593)

Campaigning 7.009 3.128 0.489 0.071 9.622 4.761 (4.914) (4.542) (2.367) (2.085) (6.178) (5.629)

Constant 4.650 18.519∗∗∗ 2.601 2.389 7.841 21.838∗∗∗ (5.931) (5.507) (2.857) (2.528) (7.456) (6.825)

Observations 239 246 239 246 239 246 R2 0.317 0.288 0.272 0.264 0.286 0.263 Adjusted R2 0.271 0.242 0.223 0.216 0.238 0.215 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

45 8 Appendix 2: Arab Barometer

Questions used:

• I will name a number of institutions, and I would like you to tell me to what extent you trust each of them (great extent=4, medium, limited, absolutely do not=1)

– [Trust in Islamists:] Muslim Brotherhood/Ennahda/Major Islamist Movement

• How often do you... (always=5, most of the time, sometimes, rarely, never=1)

– [Friday Prayers:] Attend Friday prayer

– [Frequency of Prayer:] Pray daily

• [Support for Sharia:] The government and parliament should enact laws in accordance with Islamic law (strongly disagree=1 to strongly agree=4)

• [< rights for Non-Muslims:] In a Muslim country, non-Muslims should enjoy fewer political rights than Muslims (strongly disagree=1 to strongly agree=4)

• [Men > Women in Edu:] University education for males is more important than uni- versity education for females (strongly disagree=1 to strongly agree=4)

• [Anti-US:] Do you prefer that future relations between your country and the United States become stronger, remain the same, or become weaker than they were in previous years?

• [Anti-Israel:] Which of the following statements is closest to your opinion with regard to the Palestinian question: (1) The Arab world should accept the existence of Israel as Jewish state in the Middle East only when Palestinians accept it. (2) The Arab world should not accept the existence of Israel as a Jewish state in the Middle East. (=1 if agree with second statement)

46 • [Inequality:] I am going to ask a number of questions related to the current govern- ment’s performance in specific areas. How would you evaluate the current government’s performance on... Narrowing the gap between rich and poor (very good=1, good, bad, very bad=4)

• [Corruption:] In your opinion, to what extent is the national government working to crackdown on corruption? (a large extent=1, medium, small, not at all=4)

• Arrange the following characteristics in order of their importance for a person to be qualified for political leadership in the country (1-7, 7=most important):

– [Want Integrity:] Integrity

– [Want Piety:] Piety

• [Rel not Corrupt:] Today, religious leaders are as likely to be corrupt as nonreligious leaders (strongly agree to strongly disagree)

• [Support for Democracy:] A democratic system may have problems, yet it is better than other systems (strongly disagree to strongly agree=4)

47 Table 6: Friday Prayers and Trust in Islamists (OLS)

Dependent variable: Trust in Islamists (1-4) 2011 2013 2019 (1) (2) (3) Friday Prayers 0.075∗∗∗ (0.029) 0.040∗∗∗ (0.010) 0.040∗∗∗ (0.010)

Covariates Frequency of Prayer 0.041 (0.036) 0.076∗∗∗ (0.016) 0.032∗∗∗ (0.011) Support for Sharia 0.167∗∗∗ (0.040) 0.107∗∗∗ (0.015) 0.097∗∗∗ (0.011) < rights for Non-Muslims −0.132∗∗∗ (0.039) 0.047∗∗∗ (0.014) 0.056∗∗∗ (0.015) Men > Women in Edu −0.029 (0.035) 0.017 (0.013) 0.039∗∗∗ (0.015) Anti-US −0.013 (0.038) −0.012 (0.014) 0.045∗∗∗ (0.015) Anti-Israel −0.031 (0.065) 0.071∗∗ (0.028) 0.005 (0.015) Inequality −0.053 (0.041) −0.196∗∗∗ (0.016) −0.117∗∗∗ (0.017) Corruption −0.087∗∗ (0.041) −0.210∗∗∗ (0.013) −0.046∗∗∗ (0.012) Want Integrity 0.020 (0.018) −0.002 (0.006) Want Piety 0.080∗∗∗ (0.017) 0.068∗∗∗ (0.006) Rel not Corrupt 0.120∗∗∗ (0.015) Democracy 0.027∗∗ (0.013) 0.020∗∗∗ (0.004) 0.025∗∗∗ (0.004) Age −0.007∗∗ (0.003) −0.001 (0.001) −0.001 (0.001) Female −0.151∗ (0.078) 0.031 (0.025) 0.117∗∗∗ (0.034) Married −0.034 (0.080) 0.039 (0.027) −0.040 (0.030) Education −0.060∗∗∗ (0.020) 0.011 (0.007) −0.006 (0.009) Income −0.001 (0.001) 0.0002 (0.0002) −0.016∗∗∗ (0.004) Unemployed −0.302∗∗∗ (0.107) 0.027 (0.038) −0.048 (0.043) Country FE XXX Constant 1.898∗∗∗ (0.321) 1.950∗∗∗ (0.149) 0.944∗∗∗ (0.129) Observations 1,395 7,070 5,125 R2 0.091 0.223 0.123 Adjusted R2 0.079 0.220 0.119 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

48 Table 7: Friday Prayers and Trust in Islamists, by Gender (OLS)

Dependent variable: Trust in Islamists (1-4) 2011 2013 2019 (1) (2) (3) Friday Prayers 0.079∗ (0.044) 0.064∗∗∗ (0.017) 0.046∗∗∗ (0.015) Friday Prayers*Female −0.006 (0.050) −0.035∗ (0.019) −0.011 (0.020)

Covariates Frequency of Prayer −0.128 (0.208) 0.172∗∗ (0.082) 0.154∗∗ (0.073) Support for Sharia 0.038 (0.040) 0.066∗∗∗ (0.017) 0.031∗∗∗ (0.011) < rights for Non-Muslims 0.167∗∗∗ (0.040) 0.107∗∗∗ (0.015) 0.097∗∗∗ (0.011) Men > Women in Edu −0.132∗∗∗ (0.039) 0.046∗∗∗ (0.014) 0.056∗∗∗ (0.015) Anti-US −0.029 (0.035) 0.018 (0.013) 0.039∗∗∗ (0.015) Anti-Israel −0.012 (0.038) −0.013 (0.014) 0.045∗∗∗ (0.015) Inequality −0.030 (0.065) 0.070∗∗ (0.028) 0.004 (0.015) Corruption −0.053 (0.042) −0.196∗∗∗ (0.016) −0.117∗∗∗ (0.017) Want Integrity −0.087∗∗ (0.041) −0.211∗∗∗ (0.013) −0.047∗∗∗ (0.012) Want Piety 0.020 (0.018) −0.002 (0.006) Rel not Corrupt 0.080∗∗∗ (0.017) 0.068∗∗∗ (0.006) Democracy 0.121∗∗∗ (0.015) Age 0.026∗∗ (0.013) 0.019∗∗∗ (0.004) 0.025∗∗∗ (0.004) Female −0.007∗∗ (0.003) −0.001 (0.001) −0.001 (0.001) Married −0.034 (0.080) 0.040 (0.027) −0.040 (0.030) Education −0.060∗∗∗ (0.020) 0.010 (0.007) −0.007 (0.009) Income −0.001 (0.001) 0.0002 (0.0002) −0.016∗∗∗ (0.004) Unemployed −0.302∗∗∗ (0.107) 0.028 (0.038) −0.047 (0.043) Country FE XXX Constant 1.890∗∗∗ (0.327) 1.900∗∗∗ (0.151) 0.920∗∗∗ (0.135) Observations 1,395 7,070 5,125 R2 0.091 0.223 0.123 Adjusted R2 0.078 0.220 0.119 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

49