Case Selection and Background on Ennahda

The case of Tunisia is a particularly appropriate venue to test theories of moderation. Tunisia has emerged as the single success story of the Arab Spring, an accomplishment in no small part due to Ennahda’s willingness to compromise on a number of difficult issues. As this article will show, one of the critical compromises—accepting secular democracy—was facilitated by a decade-long process of moderation among a wing of Ennahda in exile in Western capitals. With Ennahda now hailed as a model of Islamist moderation, it is critical to understand how it came to be where it is today.

Like elsewhere in the region, Islamism began in Tunisia in response to both domestic and international factors. Domestically, the overtly secularizing reforms of Tunisia’s founding father, Habib Bourguiba—who notoriously drank orange juice on national television during the fasting month of Ramadan^{Footnote 6}—unsettled the more conservative sentiments of many Tunisians. Tunisian Islamism also benefited from a broader, regional shift toward political Islam in the 1970s and 1980s following the disillusionment with Arab nationalism with the 1967 defeat to Israel and the inspiration of the Islamic Revolution in Iran. In 1981, preacher and philosophy teacher Rached Ghannouchi established the Islamic Tendency Movement (MTI). Although the Tunisian regime closed any avenues for the MTI’s formal political participation, the movement was able to operate, agitate, and proselytize on university campuses and in civil society (Wolf Reference Wolf2017).

The rise to power of President Zine El Abidine Ben Ali in 1987 initiated a brief liberalization of the regime. The MTI rebranded itself as the Ennahda (Renaissance) movement and fielded candidates on an independent list in the 1989 elections. Threatened by its popularity, Ben Ali soon cracked down on Ennahda, decimating the movement and driving its members into exile, imprisonment, or underground. For the next two decades, Ennahda would remain divided, one wing abroad in Western capitals (“Ennahda fi al-kharij”) and the other in prison or underground in Tunisia (“Ennahda fi al-dakhil”).

Following Ben Ali’s ouster in 2011, Ennahda finally reunited in Tunisia and moved quickly to prepare itself for the 2011 National Constituent Assembly (NCA) elections. Ennahda swept these elections, winning the largest share of the vote with 37% and receiving 89 of 217 seats in the assembly. Ennahda formed a “troika” government with two other parties, Congress for the Republic (CPR) and Ettakatol. In the summer of 2013, following the assassination of two liberal politicians in Tunisia and a military coup in Egypt, the troika government faced massive protests forcing them to step down from power in an agreement brokered by four civil society organizations known as the Quartet. Ennahda remains a powerful political force, helping to form a coalition government in 2015 and lead one today.

During the troika government’s tenure (2011–14), Ennahda made a number of critical compromises in accepting the principles of secular democracy. The first, perhaps the most important, was to not include in the 2014 constitution a reference to shari‘a, or Islamic law, as the basis of legislation. Such a clause would have directly limited popular sovereignty in crafting legislation, contradicting democratic norms. Early in the transition, Rached Ghannouchi, the head of Ennahda who had been exiled in London, assured Tunisia’s secularists and the international community that the first article of the constitution—that the official language is Arabic and religion is Islam—would be left as is; “there will be no other references to religion in the constitution.”^{Footnote 7}

Yet that campaign promise did not sit well with some in Ennahda. In the spring of 2012, conservative firebrands Habib Ellouze and Sadok Chourou, both of whom had spent much of the last two decades in Tunisian prisons, forced an internal debate on the matter. The debate was protracted, reinforcing fears among Tunisia’s secularists that Ennahda harbored a hidden, fundamentalist agenda (Marks Reference Marks2014, 20–2). Although Ennahda’s conservative wing lost that round, they had not given up. When it came time to draft the new constitution, they introduced amendment 42 to article 1, which would have added to the constitution’s first article: “the Quran and the Sunna are the main sources of its [Tunisia’s] legislation.” If a reference to the shari‘a had been ruled out, perhaps a reference to the sources of the shari‘a—the Quran and the Sunna—could gain enough votes.

When it came up for a vote on January 4, 2014, Ennahda MPs were uncharacteristically divided (Figure 1).^{Footnote 8} Twenty-two of 89 Ennahda MPs (25%) voted for this amendment, whereas 38 stuck to the party line, which was to abstain. Twelve were absent during the vote, whereas another 17 voted no. Although the amendment ultimately failed, it generated important and illuminating internal disagreement that will permit us to separate Ennahda MPs into more secular and more Islamist on the issue of popular v. divine sovereignty.

FIGURE 1.

Distribution of Ennahda Votes, Quran, and Sunna as Basis of Legislation January 4, 2014

Additional controversies revolved around religious freedom, the second pillar of secularism. Amendment 62 of article 6, voted down on January 4, 2014,^{Footnote 9} would have eliminated the constitution’s guarantee to freedom of conscience, commonly interpreted as permitting atheism. As MPs were debating this amendment, Ennahda MP Habib Ellouze called secular MP Mongi Rahoui an “enemy of Islam,” prompting secularists to introduce a new amendment to article 6 prohibiting takfir—the labeling of another Muslim as an apostate, a religious incitement to violence.^{Footnote 10} This amendment passed on January 5, 2014.^{Footnote 11} The Ennahda bloc was considerably divided on both of these votes as well (see online appendix).

In short, Tunisia’s National Constituent Assembly (NCA), tasked with drafting the 2014 constitution, was the site of major compromises, including on secular-religious issues. Why did some Ennahda MPs vote in favor of these secular compromises, whereas others did not? In particular, did the Ennahda MPs who had lived in secular democracies vote more secularly?

Dependent Variables

I pursue two strategies to separate Ennahda MPs into “more secular” and “more Islamist.” The first is targeted: to examine just the three major compromises highlighted above, including (1) whether to make the Quran and Sunna the basis of legislation, (2) whether to allow freedom of conscience, and (3) whether to ban the practice of takfir. Case knowledge suggests that these were three of most divisive compromises, and all directly relate to the secular-religious cleavage. I also present a principal components analysis (PCA) combining these three votes.^{Footnote 12}

While these three votes allow us to hone in on the primary issue of interest, they may be biased by idiosyncrasies of these particular votes. The second approach is therefore more comprehensive: to analyze all 1,731 roll-call votes in the NCA to extract an underlying ideal point, or ideology, for each Ennahda MP.^{Footnote 13} The ideal point approach is similar to that which is used in analyzing the U.S. Congress, and allows us to separate Ennahda MPs as more liberal or more conservative based on their entire voting records. The limitation is that the separation is influenced not only by votes related to secularism, but also other divisive issues for Ennahda, such as transitional justice.^{Footnote 14} The individual votes and the ideal point analyses are therefore different but complementary approaches to measuring MPs’ ideologies.

For both analyses, my preferred approach is to code an ordinal variable with three voting options: yes, abstain, or no (with absences coded as missing data). While many ideal point models treat abstentions as missing data as well, in our case they are quite informative. Rather than simply meaning “don’t know,” abstentions are meant to signal that: “I don’t fully agree with the position of my parliamentary group,” explained one Ennahda MP. “An abstention sends a message to my parliamentary group that maybe we should discuss a little more about this position.”^{Footnote 15} Ennahda MP Osama al-Saghir concurred, arguing that an abstention means that “I like the idea but the article as written is not as I think it should be.”^{Footnote 16} At least in the Tunisian context, abstentions register an intermediary vote between yes and no, and thus provide important information about MPs’ preferences.

For the individual votes (approach 1), I therefore code those MPs voting secularly—i.e., “no” to the Quran/Sunna, “no” to removing freedom of conscience, and “yes” to banning takfir—as a 1. Those voting Islamist are coded as −1, and those abstaining as a 0. As the dependent variable is ordinal, I run ordered logistic regressions using the lrm R function.

In conducting the ideal point analysis (approach 2), I use the ordinal Item Response Theory (ordIRT) model developed by Imai, Lo, and Olmsted (Reference Imai, Lo and Olmsted2016) that permits abstentions as a middle category between yes and no. However, as a robustness check, I also employ the more standard binary model that excludes abstentions, using the “w-nominate” package (Poole et al. Reference Poole, Lewis, Lo and Carroll2011). Unlike ordIRT, w-nominate allows the user to separate MPs along multiple dimensions, and I extract the first dimension on the assumption that the secular-religious cleavage was the primary cleavage during this time (Berman and Nugent Reference Berman and Nugent2015; Brownlee, Masoud, and Reynolds Reference Brownlee, Masoud and Reynolds2015).

The Cost of Voting

The use of parliamentary votes to infer Islamists’ preferences is a break from the literature, which instead measures ideology through public or private statements, such as a party platform, press statement, or private interview with the researcher.^{Footnote 17} Given the relatively costless nature of these statements, however, one potential problem is preference falsification: especially when Islamists are speaking with the West, they may have incentives to misrepresent their true beliefs to appear more moderate, or selectively omit beliefs that would make them appear hostile. These sources may therefore provide little guidance for how Islamists will actually behave upon coming to power.

These concerns are mitigated when examining parliamentary votes. Votes are not only an actual measure of what Islamists do once in power but they are also costly. Ennahda has maintained the highest party discipline in Tunisia; its MPs face considerable social and organizational pressure to toe the party line. Divergences from the party line carry real consequences, including a lower likelihood of being renominated by the party to run in subsequent elections. They also have a direct, substantive impact on the constitution, laws, and future of the country. As a result, divergences from the party line, whether in a more liberal or conservative direction, likely reflect an MP’s strongly held beliefs and are unlikely to be preference falsification.

To provide evidence of one such cost, I examine whether those who voted against the Ennahda party line on any of the three votes outlined above were less likely to be renominated by the party to run in the 2014 elections. Divergences from the party line were coded as either voting “yes” to the Quran and Sunna as the basis of legislation, “yes” to removing freedom of conscience, or “no” to banning takfir.

Figure 2 finds strong evidence for this proposition. Of the 86 MPs who chose to remain within Ennahda, those who diverged from the party line were 30% less likely to be renominated than those who stuck to the party line (p = 0.0066), even when controlling for a number of demographics (see appendix). Similar results obtain for the ideal point analysis, where more conservative MPs were significantly less likely to be renominated (p = 0.00164, appendix). An interview with Ennahda executive board member and MP Sahbi Atig confirmed that although party lists were drawn up by each governorate level office of Ennahda, the executive board “intervened in four or five governorates” to block certain individuals. Atig in particular mentioned the district of Sfax 2, where the aforementioned MP Habib Ellouz was not renominated.^{Footnote 18}

FIGURE 2.

Likelihood of Renomination by Votes on Secularism

Interviews also revealed that MPs knew about this cost ahead of time. In the lead-up to controversial votes, for instance, the Ennahda headquarters had individually called those MPs leaning toward voting against the party line to inform them that they may not be renominated if they do so. MP Mohamed Saidi recalled that: “We were called, we were even threatened, if you vote for this you are killing your political future!”^{Footnote 19}

Given this cost of voting, diverging from the party line likely reflects an MP’s strongly held convictions and preferences, and is unlikely to be cheap talk or preference falsification. Parliamentary votes should thus provide a fairly accurate measure of moderation.

Independent Variables and Descriptive Statistics

Based on the theory of secular diffusion, I hypothesize that Ennahda MPs who studied or were exiled in secular democracies would vote more secularly in the NCA than their counterparts who had not. Data for time spent in secular democracies come from public biographies of the MPs. Each MP submitted a detailed biography to al-Bawsala, a watchdog organization that monitored the parliament.^{Footnote 20} Each biography was cross-checked with other online sources as well as with the Ennahda headquarters.^{Footnote 21}

Of the 89 Ennahda MPs, 18 had studied or were exiled in secular democracies in the West, including Belgium (1), Canada (1), France (12), Germany (1), Italy (2), and the United States (1). These 18 MPs spent an average of fifteen years abroad, ranging from two to thirty years (see histogram in online appendix), and occurring primarily in the 1990s and 2000s. In addition, nine other MPs had studied or been exiled in other authoritarian regimes in the region (Algeria, Morocco, Mauritania, Qatar, and Sudan). These MPs can serve as a placebo demonstrating that it is not traveling abroad that is important but the destination country in particular. The remaining 62 MPs had only lived in Tunisia.

Descriptive statistics provide initial support for the theory. Figure 3 plots the votes on whether the Quran and the Sunna should be a basis of legislation for MPs who have spent time in secular democracies in the West (right), lived in autocracies in the Middle East and North Africa (left), or lived only in Tunisia (middle). Of the 16 who lived in the West and were present for this vote, only one (6%) voted for the Quran and the Sunna to be the basis of legislation, whereas the remaining 15 either voted no or abstained. By contrast, those who only lived in Tunisia (middle) were more conservative, with 32% voting for the Quran and the Sunna. Finally, those who had lived in other MENA countries were the most conservative, with 57% voting for the Quran and the Sunna. In other words, Ennahda MPs who had lived in the West were about 25 percentage points less likely to vote for the Quran and Sunna to be the basis of legislation than their counterparts who had lived only in Tunisia, a difference which a t-test reveals is statistically significant at p = 0.007.

FIGURE 3.

Descriptive Statistics, Quran, and Sunna

Similar results obtain for the other two secularism votes. Not a single Ennahda MP who lived in the West voted to remove freedom of conscience, whereas 22% of those who never left Tunisia did so (p < 0.000). Meanwhile, 89% of those who lived in the West voted to prohibit takfir, compared with only 66% for their colleagues at home (p = 0.06).

The ideal point analysis mirrors these results. Figure 4 plots the ideal points of the 89 Ennahda MPs from most liberal (left) to most conservative (right).^{Footnote 22} As can be seen, Ennahda MPs who lived in the West (names in blue) are visually more concentrated toward the left than right, with a mean of −0.48, compared with −0.04 for those who lived only in Tunisia (red) and 0.56 for those who lived in other MENA countries (green).

FIGURE 4.

Ideal Points of Ennahda MPs (Left = Liberal and Right = Conservative)

For the purposes of visualization, these descriptive plots presented a simple dichotomous variable of whether an Islamist MP had lived in a secular democracy. However, there is reason to believe that the length of time one stayed abroad may matter as well. On average, the longer an Islamist MP lived abroad, the more likely he or she is to become secular. Accordingly, in the regression models below, I use the length of time, presenting the dichotomous variable in the appendix. The length of time is standardized by dividing by two standard deviations (Gelman Reference Gelman2008).

Control Variables

The introduction of regression analysis to studies of moderation allows us to isolate the link between living in the West and moderation while controlling for a number of covariates. For each dependent variable, I present two models. The first controls only for demographic characteristics: age, gender, level of education, whether they have a theology degree, as well as fixed effects for hometown (coded as coast, southwest, southeast, interior, or northwest) and prior occupation (professor, teacher, lawyer, scientist, business, doctor, government, other, or none).

The second model then adds several variables capturing the existing theories of Islamist moderation outlined in the literature review. To capture the interactions with other political parties that Schwedler (Reference Schwedler2006) highlights in the inclusion-moderation hypothesis, I control for activism, coded as participation in student unions or professional syndicates. To account for the repression that is also thought to breed moderation, I control for whether the MP had ever been in prison. Nugent (Reference Nugent2017), for instance, argues that since Ennahda and secular politicians had both been repressed, their joint experience in jail may breed ideological convergence, moderating Ennahda MPs who had been imprisoned. Finally, to capture the political incentives MPs might have to moderate to appeal to the median voter, I include three additional variables:

1. 1. Ennahda Vote Share: MPs may vote more secular in those districts where Ennahda was weak. Where Ennahda’s vote share in 2011 was low, it would need to reach out to a larger constituency to win reelection, and thus would have more incentives to moderate.

2. 2. CPR Vote Share: Of all the secular parties, the Congress for the Republic (CPR) was the most friendly to Islamism, and indeed included some former Islamists like Imed Daimi among its leadership. CPR’s success in a district may therefore signal a friendly constituency for Islamism even beyond Ennahda’s base. CPR vote share should therefore be correlated with less secularism.

3. 3. Place on Party List (inverted): The lower an MP is on Ennahda’s party list, the more incentive he or she may have to stick to the party line to advance up the ladder. The opposite effect could also occur: Ennahda may have placed them lower down the list precisely because they were known to be less loyal to the party line.^{Footnote 23} I invert the list so that regardless of the number of seats Ennahda won in a governorate, those at the bottom receive the same low value (1).^{Footnote 24}

4. 4. Electoral District: Finally, I include a dummy variable indicating whether the MP represents a district in the coast or abroad (1), rather than in Tunisia’s interior regions (0), on the assumption that these districts are likely to be more secular. Each of the eight Ennahda MPs representing constituencies abroad (Italy, France, Germany, Americas, etc.) had themselves traveled abroad. Including this control thus demonstrates that the effect of the West is not driven by those representing constituencies abroad.^{Footnote 25}

Several of these variables occur posttreatment, and thus the effect of living in the West may be running through them. For instance, Ennahda may have placed MPs who lived abroad higher (or lower) on the party list or in governorates that were competitive (or uncompetitive). In short, because some of these variables occur “posttreatment,” we may expect the effect of living in the West to weaken; yet, it is important to test the robustness of the results to these alternative, electoral incentive explanations.

Results

Table 1 presents the first dependent variable: the individual secularism votes, including on the Quran/Sunna as the basis of legislation (models 1–2), freedom of conscience (3–4), the prohibition of takfir (5–6), and the principal components analysis of the three (7–8). For each dependent variable, positive values indicate more secular votes.

TABLE 1.

Secular Diffusion and Individual Secularism Votes

Note: *p < 0.1; **p < 0.05; ***p < 0.01.

The coefficient of interest is the time spent in the West, which should be positive if the secular diffusion thesis is correct. In all eight models, the effect of living in the West is positive and statistically significant (p < 0.05 or p < 0.01). MPs who had lived in the West voted significantly more secularly, even when controlling for demographic variables and alternative explanations.

Included in all models was also the placebo: those Ennahda MPs who also lived abroad, but in other (authoritarian) Middle Eastern and North African countries. Those elsewhere in the MENA region were no more secular than those who remained in Tunisia, and on the contrary were leaning more conservative.^{Footnote 26} It is not simply being abroad, but being abroad in secular democracies, that is correlated with secularism.

The electoral incentives also perform well in predicting these secularism votes. Ennahda MPs representing districts where Ennahda and CPR did well, signaling a relatively more Islamist-friendly constituency, voted more Islamist, lending credence to the median voter theorem at the governorate level. Place on the party list also reached significance, with MPs higher on the list voting more secularly.

Other variables attain significance for some votes but not others. Women were more likely to vote against the Quran and Sunna as the basis of legislation, whereas higher educated MPs were less likely to ban takfir. Activism, representing MPs who had been involved in student unions or professional syndicates, was correlated with support for freedom of conscience and banning takfir, but not for the other models. Imprisonment had no effect on these votes.

Table 2 presents the results for the second approach: the ideal points that were extracted from all 1,731 votes in the assembly. The sign is flipped from Figure 4, with positive values now indicating more liberal ideal points. Models 1–2 employ the ordinal ideal point analysis, whereas 3–4 employ the binary version (excluding abstentions).

TABLE 2.

Secular Diffusion among Ennahda MPs (OLS)

Note: *p < 0.1; **p < 0.05; ***p < 0.01.

In all four models, the effect of living in the West is again positive and statistically significant (p < 0.05). MPs who had lived in the West held significantly more liberal ideal points, even when controlling for demographic variables and alternative explanations. Overall, having lived in a secular democracy appears to be one of if not the most important predictor of liberal voting records.

For the ideal point analysis, few of the covariates were significant. The only consistently significant covariate was place on the party list, with MPs higher up on the list having more liberal ideal points. Prior activism was correlated with more liberal ideal points in the ordinal model but not in the binary. Imprisonment again had no significant effect. In short, the existing explanations for Islamist moderation—electoral incentives, repression, and interactions with other parties—do a mixed job at best in predicting secularism, at least as measured herein. These results of course do not rule out the potential importance of these factors in moderating Ennahda MPs in other issue areas, for instance, on women’s rights or lustration, but appear to have little effect on breeding an acceptance of secularism.

Tables 5–6 (appendix) demonstrate that results hold when using the dichotomous West variable rather than the length of time spent abroad. Tables 7–8 experiment with separating those who traveled to the West for education from those fleeing repression. Theoretically, the mechanisms should operate for both, although mechanism 1 may be stronger among those fleeing repression (and thus whose personal situation arguably improved more). While the distinction is not perfect (6 MPs did both), we can add both education and exile into the regression. Results are inconclusive: for the ideal point analysis, it is those seeking exile who are driving the secularizing effect of living in the West. However, for the individual secularism votes, those seeking education claim the stronger effect, including significant effects for banning takfir and for the PCA.

In sum, one of the strongest and most consistent predictors of secularism among Ennahda MPs was living in secular democracies, whether for education or exile.

Alternative Explanations

Before demonstrating support for my hypothesized mechanisms, it is important to first address two alternative explanations. First, perhaps Ennahda MPs who had lived in the West became more secular not because of any secular diffusion from these countries but rather because being in the West facilitated meetings and close relationships with Ennahda leader Rached Ghannouchi, who was exiled in London. As a result, these MPs may be voting more secular—i.e., in the way Ghannouchi, as head of the party, wants them to vote—because Ghannouchi has greater personal influence over them.^{Footnote 27} This explanation, however, would imply that these MPs should across the board vote closer to Ghannouchi and the party line than other MPs, not just on votes related to secularism. To test for this explanation, we can calculate each MP’s overall propensity to diverge from the party line using all 1,731 votes in the NCA. I calculate the party line for each vote as the modal vote (yes, no, or abstain) on each bill. In other words, rather than calculating who is more liberal and more conservative, we can examine who is most likely to diverge from the party line whether in a liberal or conservative direction.

I find that Ennahda MPs who had lived in the West did not vote significantly closer to the party line than MPs who had remained in Tunisia (see figure in appendix). On average, Ennahda MPs broke with the party line 13% of the time. MPs who had lived in the West voted against the party line 12.1% of the time, whereas MPs who lived only in Tunisia did so at 13.7%. This difference is not statistically significant (p = 0.24). We can therefore reject the alternative explanation that Ennahda MPs who lived in the West happened to be closer to the party line or to Ghannouchi than other MPs.^{Footnote 28}

A second and potentially more damaging alternative explanation would be selection bias: perhaps those Ennahda MPs who lived in the West chose to do so because they were already more secular—it was not that they became more secular, but rather they already were. Unfortunately, there are no data on these MPs’ ideological preferences before going abroad to be able to directly test for selection bias. Although we are not able to rule out endogeneity entirely, allow me to address this concern in three ways, each of which has its flaws but together reveal a consistent story.

The first approach is to examine if certain covariates that we expect to be correlated with secularism are associated with going to the West. For instance, we may think that age or education may be correlated with secularism. Similarly, the results above suggest that women may be more secular, whereas in terms of occupation, professors/teachers were leaning more secular.

Beyond these individual-level covariates, we can examine if those who traveled to the West were from more secular districts. We can measure the religiosity of each district in three ways: First, we can use Ennahda’s performance in the 1989 elections, the one time that it was permitted to contest elections under autocracy.^{Footnote 29} This variable ranges from 0 in multiple districts to 30% in Tunis 2. Second, we can examine the number of mosques per capita in each governorate,^{Footnote 30} a variable that ranges from 1.9 mosques per 10,000 citizens in Gabes to 11.6 in Monastir. Finally, we can exploit Tunisia’s regional divisions, as citizens from the wealthy coastal regions tend to be more secular (Berman and Nugent Reference Berman and Nugent2015) than their counterparts in the impoverished interior, Northwestern, or Southern regions (Boughzala and Hamdi Reference Boughzala and Hamdi2014).

Table 3, however, does not show evidence of any these selection effects. Ennahda MPs who had lived in the West were roughly the same age, had the same level of education, and were not significantly different in their occupations than those who stayed in Tunisia. Moreover, they were no more likely to hail from the coast or from less religious districts. Where there is imbalance—gender—the results actually cut in the opposite direction. Those who went abroad were more likely to be male and thus less secular. Although these are admittedly crude proxies for secularism, they suggest that the MPs who lived in the West were not demographically likely to already be more secular.^{Footnote 31}

TABLE 3.

Covariate Balance among Ennahda MPs

A second approach is to subset the data to a group where going abroad was more plausibly uncorrelated with secularism. In particular, we can exploit the massive political crackdown on Ennahda in 1991, in which all members of Ennahda, whether more secular or more Islamist, faced the prospect of arrest. In a private interview with the author, the director of internal security within the military from 1988–2000 noted that the 1991 crackdown was wide-reaching and indiscriminate, targeting leaders, members, and even sympathizers of Ennahda.^{Footnote 32} Ennahda claims that 30,000 of its members were imprisoned during this period—the largest crackdown in Ennahda’s history.^{Footnote 33}

Whether an Ennahda member managed to escape arrest during the 1991 crackdown was primarily a matter of chance. Interviews with several of Ennahda MPs are illustrative in this regard. Ennahda MP Mohamed Zrig, who managed to escape from his hometown in Gabes in southern Tunisia to Canada, explained:

MP Walid Bennani, from the interior town of Kasserine, who managed to escape to Belgium, likewise attributes his escape to chance:

These interviews suggest that whether an Ennahda MP managed to flee or was arrested in 1991 was unrelated to their level of secularism: all were targeted for arrest. To conduct this analysis, I will subset to MPs who were at least 18 years old and in Tunisia in 1991 (78 of 89 MPs), and therefore could have been arrested. The primary independent variable here, Exile in 1991, will lump together all MPs who fled the country, whether going to the West or MENA.^{Footnote 36} Although escaping Tunisia is plausibly uncorrelated with secularism, where they settled is likely not.

Table 9 (appendix) presents the results. Across all six dependent variables—the two ideal point analyses, the three individual votes, and the PCA—MPs who had been exiled in 1991 voted significantly more secularly than those who remained in Tunisia. This finding is particularly surprising, given that now the reference category includes several MPs who had gone abroad before 1991.

The final piece of evidence against selection bias is the strongest: Ennahda MPs themselves admit that their time abroad had a causal impact on their beliefs. Thus, even if there was some selection bias in who went abroad, there was at least an additional effect of living in the West on their preferences. Although we will hear from more MPs who had gone abroad in the mechanisms section, for now, it will suffice to hear from Mohamed Zrig, who spent 20 years in Canada:

Even those who were imprisoned recognize this ideological moderation among those exiled in the West.^{Footnote 38} Conservative Ennahda MP Sadok Chourou, who spent almost two decades in prison, affirmed:

Ennahda MP Habib Ellouze, who also spent two decades in prison, concurred:

In sum, even if there was an initial selection effect in who went abroad, Ennahda itself recognizes that its experiences in exile have at least had an additional effect. To see why, and to explore the mechanisms by which this effect occurred, I turn next to several interviews with these Ennahda MPs.