From Nomadic Traditionalists to Sedentary Scripturalists

Religion and ethnicity are inextricably linked in discourse within and about Central Asia. One common narrative suggests that as a result of differences between historically sedentary and nomadic populations, ethnic Tajiks and Uzbeks are naturally more religious and more likely to radicalize than their Kazakh and Kyrgyz neighbors. Using extensive data available from the Pew Research Center’s 2012 The World’s Muslims survey, this study examines whether such claims stand up to empirical scrutiny. The data cast doubt on simplified versions of this discourse and suggest that future analyses should focus attention on individual-level explanations rather than potentially essentializing group-based narratives.

that by virtue of their religiosity, Tajiks and Uzbeks are more likely to radicalize and actively support violence in the name of Islam-is even more worrisome, as it is often used to justify state monitoring and repression of these groups. Systematic evidence for both of these claims is also sparse. The first is most commonly supported by observational anecdotes wherein piety is assessed on the basis of a single or narrow range of characteristic(s), such as frequency of prayer or mosque attendance. The validity of the second claim typically relies on references to a very small number of high-profile cases of violence in the region as well as arrest records of so-called "extremists," records that are notorious for their bias.
Deploying a rich set of survey data recently released by the Pew Research Center, this study seeks to explore the validity of the ethnicity-religiosity-radicalism nexus in greater depth. Drawing on 4,708 face-to-face interviews with self-identified Muslims in Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan, Pew's 2012 The World's Muslims dataset represents the largest-scale cross-country survey undertaken in the region that contains a multitude of questions concerning religious practices and beliefs along with data on ethnicity. I carefully unpack the survey data and subject it to close examination, employing a series of ANOVA tests to understand whether ethnic differences are indeed associated with variations in religious observance and beliefs. The data reveal that the ties between religiosity, ethnicity, and radicalism are much more complex than the standard discourse at times leads us to believe. Using a number of different measures for both religiosity and radicalism, it becomes clear that whether the ethno-religious narrative holds up to empirical scrutiny depends on how one conceptualizes and assesses both religiosity and radicalism. In some instances Tajiks and Uzbeks do indeed prove more religious and less moderate than their historically nomadic neighbors, but they are not axiomatically so. In fact, when looking at the most stringent measures of radicalism-particularly whether one supports violence to "defend" Islam-ethnic Kyrgyz prove most "extreme." I subject this surprising finding to further scrutiny using logistic regression analyses and find that, after controlling for a number of additional variables, ethnicity shows no impact on one's degree of radicalism. Instead, support for the use of violence is correlated with individual-level factors tied to one's worldview as well as grievances about political governance and personal welfare. Taken together, these findings suggest that scholars and policymakers should be cautious about forwarding simplified ethno-religious narratives to explain contemporary religious trends in the region.
In the pages that follow, I first offer an overview of the two claims that lie at the heart of the ethno-religious narrative as well as some of the problems with these claims as highlighted by previous scholarship. I then analyze the survey data, beginning with comparisons of the results by ethnicity, followed by a more complex analysis that examines the effects of ethnicity alongside a host of other variables that may impact upon radicalism. I conclude by discussing the broader implications of these findings.

COMPARING NOMADIC TRADITIONALISTS TO SEDENTARY SCRIPTURALISTS
Olivier Roy (2000) writes that "one can speak of two variants of Islam, corresponding to an opposition … between tribal zones (Kazakhs, Turkmen and Kyrgyz) and areas of longstanding urban civilizations that were Islamicized after the Arab conquest (Tajiks and Uzbeks from Transoxiania)." Islam, he suggests, "was late in being imposed" in the "tribal zones" and to this day "incorporates elements deriving from the shamanistic traditions of Turkic nomads." On the other hand, Roy argues, "The Islam of Transoxiania is a product of the madrasas … of Samarkand and Bukhara," and it is "often a fundamentalist Islam" . This, in its essence, is the dichotomized picture of Islam portrayed in much of the literature on, and the public discourse prevalent in, Central Asia. Islam was accepted earlier, in the seventh-ninth centuries, and "more readily," by sedentary populations, "particularly the ancestors of today's Uzbeks and Tajiks" (Walker 2003, 23); and though the nomads of Kazakh, Kyrgyz, and Turkmen ancestry had contact with Muslims in earlier historical periods, their mobility prevented them from becoming fully "Islamicized" until sometime between the seventeenth and nineteenth centuries. This, in turn, precluded their faith and traditions from developing "the same depth" that was characteristic of their settled neighbors (Ro'i 2003, 242). Moreover, because they lacked the formal institutions of settled and urbanized areas-institutions such as mosques and madrasas-and the authoritative pronouncements that could be issued from such institutions, the nomads more easily mixed the new Islamic rites, practices, and beliefs with those of their "traditional" shamanism (Khalid 2007, 33;Omelicheva 2010, 175;2011, 245;Imart 1986).
This narrative is also frequently voiced by Central Asian officials. Take then Kyrgyzstani president Askar Akaev's words, for example: Here in Kyrgyzstan Islam was assimilated in a rather untraditional form. What we see here are the outward trappings of Islam without the exalted religious fanaticism and ideology. … Our brand of Islam absorbed many of the cultural traditions of the peoples in the region. (Quoted in McBrien and Pelkmans 2008, 91) Indeed, the notion that the particular form of practice of ethnic Kyrgyz is more culturally than scripturally authentic is often taken for granted by members of the public (Louw 2012(Louw , 2013McBrien and Pelkmans 2008), and average Kyrgyz citizens explicitly embrace the ethnic dichotomy drawn between themselves and others, particularly Uzbeks. As Chris Hann and Mathijs Pelkmans (2009) observe, "In everyday speech, Kyrgyz described their Uzbek neighbors as 'more Muslim' and attributed their own less dogmatic brand of faith to their nomadic past" (1528). Some even describe fellow Kyrgyz whom they consider too religious as being "like Uzbeks" (Tromble 2014;Flynn and Kosmarskaya 2012, 460). Similarly, Pawel Jessa (2006) notes the "widely held view" in Kazakhstan "that Islam has never penetrated the Kazakh spirit" (174) and points out that even members of the muftiate seem to "endorse the familiar dichotomy between scriptural Islam and practices … commonly glossed as 'popular Islam'" (171).
In fact, public officials' policies and pronouncements in Uzbekistan and Tajikistan also embrace this dichotomy, laying claim to the "popular" or "traditional" practices for their own, historically sedentary populations. The policies of Uzbekistan's late president, Islam Karimov, are particularly well known in this regard: President Karimov promotes a "cultural" or "civil" Islam, emphasizing the tradition of religious tolerance, which is presented as a special characteristic of the local interpretation of Islam. … Local practices such as shrine veneration and Sufism are integral parts of the official definition of a nationalized, "Uzbek" Islam. … Through public speeches, the state-controlled media, slogans on billboards, the teaching of subjects such as "spirituality and enlightenment" in schools, and the sending of university professors on agitation tours into villages, the government spreads its view of what constitutes "an Uzbek" and "Uzbek Muslimness." (Hilgers 2006, 83; see also Rasanayagam 2011) NOMADIC TRADITIONALISTS AND SEDENTARY SCRIPTURALISTS The ethno-religious narrative has thus been instrumentalized by state actors who seek to underscore and enforce a "moderate" and "safe" vision of Islamic practice in the region.
Such instrumentalization should already make us wary of this ethnicized discourse. However, as many historians have observed, there are other reasons to doubt its authenticity. In chronicling the development of Kazakh society under tsarist rule, Robert Crews (2006, 192-240) ties this narrative to the orientalist agendas of the region's colonizers. Among those who believed Islam could serve as a "civilizing" force for the tsarist empire, positive reports were given of the "orthodox" characteristics developing and flourishing among Kazakhs. Yet among those who saw Islam as dangerous and subversive, much was made of the notion that Islam was "unnatural" to Kazakhs. Indeed, some of these latter observers (especially ethnographers) simply rejected Kazakhs' own proclamations of Muslim identity. Devin DeWeese (1994) similarly argues that "a persistent fear and hostility toward Islam … combined with general unfamiliarity with the Inner Asian world" (4) has led both historical and contemporary observers to "uncritically accept" the notion that "Islam 'sat lightly' upon the Inner Asian nomad, whose 'conversion' was in name only" (9). And Bruce Privratsky (2001, 10) suggests that contemporary Kazakhs have come to view themselves as less Muslim than their sedentarized neighbors "in the light not so much of their own religious experience, but of the ethnographic literature" and journalistic writings about them.
Though the ethno-religious narrative has long historical roots in Central Asia, many note that this account also served the Soviet view of religion, which sought to dismiss Islam "as simply an 'ideology,' as an intellectual system used by the ruling class to maintain its class privileges and hegemony over the working class" (Frank 2001, 8). And as DeWeese (2002,310) observes, the Sovietological approach, in turn, was rooted in "disdain both for 'excessive' ritual and for superstitious 'folk religion'" and "served the purpose of fragmenting the traditional life of Muslims into components that could be labelled 'real Islam,' and others that could be relegated to various categories ranging from 'survivals from paganism' … to 'degenerate' or 'popular' Islam" (see also Saroyan 1997).
Of course, such arguments do not disprove the empirical existence of a historical ethno-religious dichotomy. Indeed, there is strong evidence that the number of mosques, Islamic officials, and Islamic schools was considerably higher in the settled regions of Central Asia (Bazarov 1997). However, while such data confirm that there were more religious institutions in settled areas in earlier periods, they say little about the type of observance, practice, and belief among the sedentary versus nomadic populations at these times. Yet a great deal of contemporary research does offer rich evidence of the depth and piety of faith present among historically nomadic populations in more recent years. On the one hand, this research shows that members of these ethnic groups are indeed taking up "orthodox" approaches to Islam. Alima Bissenova (2005, 255), for instance, writes of young Kazakhs studying at Al-Azhar University in Cairo-facing great hardship in an often unforgiving environment because they believe that the skills and knowledge they acquire at this prestigious institution will "confer a certain 'cultural capital' upon them back home." And Wendell Schwab (2014Schwab ( , 2015 chronicles Kazakhstan's piety movement, which discourages visits to shrines, emphasizes the importance of "avoiding polytheism (shirk) and innovation (bighdat)," and encourages "observance of daily ritual prayer, belief that there is only one true interpretation of Islam, pedagogy focused on engagement with the Qur'an and hadiths, and modest dress such as headscarves for women" (2015,255).
And yet recent research also underscores the essentialism inherent in a dichotomized religious narrative that insists on such evidence for demonstration of "real" faith among Kazakhs and Kyrgyz. David Montgomery's (2015) research in Kyrgyzstan, for example, describes the gravity with which Kyrgyz consider questions of "proper" faith and practice, as they seek to understand and fulfill "Islam's moral claims and vision" (49). And Maria Louw's (2012Louw's ( , 2013 research highlights the deep spirituality and sense of oneness with God experienced by Kyrgyz who feel intensely connected to their faith, yet themselves embrace the distinction between their more cultural, less "pure" forms of practice and "excessive" brands of Islam. Such practitioners know their faith as real and profound but argue that its "traditional" elements ward against dangerous, "extremist" beliefs.
It is this concern for, and fear of, extremism that brings us to the second core claim in the Central Asian ethnoreligious discourse.

RADICALISM IN THE ETHNO-RELIGIOUS NARRATIVE
When considering the possibilities for "extremism" or "radicalism" in Central Asia, most scholars are quick to reassure that extremists are relatively rare, that a grounding in Sufism and the Hanafi school of Islamic jurisprudence is likely to keep most Muslims in the region from radicalizing (Akiner 2003;Walker 2003). What is more, traditionalism is said to serve as a natural "barrier to the growth of" extremism (Akbarzadeh 2001, 460). In other words, close connections to folkloric traditions and "superficial" understandings of Islam make it especially difficult for radical ideas and practices to take hold. Yet there is a problematic implication built into this type of reassurance. If the embrace of traditionalism prevents extremism, those who are more likely to eschew traditionalism-those who hold scripturalist views and are more deeply committed to their faith-must also be more likely to radicalize. Thus, Mariya Omelicheva (2010) suggests that Tajiks' and Uzbeks' "prior socialization experiences with Islam"-their "early adoption of Islam and the close 358 TROMBLE observance of Islamic laws and prohibitions"-"exposed Uzbek and Tajik Muslims," more than the historically nomadic groups, "to radical Islamic teachings" (175). Ahmed Rashid (2002) similarly argues that "even today the nomadic Kazakh, Kyrgyz, and Turkmen tribes are far less Islamicized-and much less susceptible to Islamic radicalism-than their ethnic counterparts in the settled oasis areas" (29). In short, the greatest potential for radicalization-even if it is, on the whole, quite small-lies with ethnic Tajiks and Uzbeks.
This claim rests on the rather problematic assumption that religiosity itself breeds extremism, that those who are most observant are most susceptible to radicalization. John Heathershaw and David Montgomery (2014) chronicle the use of this assumption in Central Asian policy analysis and tie it to the broader narrative common in international discourse since September 11, 2001. Indeed, as Western countries have sought "to elaborate and explain the process by which individuals become radicalized to extremist violence," they have "given precedence to religion (specifically Islam) as a key factor in the radicalization process" (Aly and Streigher 2012, 849). Kundnani (2012, 9) elucidates the rationale for such thinking: For some, the question of religious belief … is most significant. If a set of religious beliefs can be identified that terrorists share with a wider group of radicals, but which "moderate" Muslims reject, then a model can be developed in which such beliefs are seen as "indicators" of radicalization.
Such a model, however, has yet to be discovered. And volumes of existing research cast doubt on the notion that increased religiosity or piety in turn increases the probability of supporting or carrying out terrorist violence (Berger 2014;Fair et al. 2012;Pisoiu 2013;Tessler and Robbins 2007;Zhirkov et al. 2014).
Still, the assumption continues to drive policy, both in the international arena and in Central Asia specifically. In Central Asia, outward signs of piety, including growing long beards or wearing the hijab or niqab, are regarded with suspicion by state authorities, and those exhibiting such characteristics run the risk of "summary arrest," long-term detainment, and even torture (Akiner 2003, 102-3; see also IHFHR and Memorial 2006, 6;Khalid 2003, 589).

RESEARCH DESIGN
With such serious consequences, the two claims explicated here deserve greater empirical scrutiny. As Shirin Akiner (2003) observes, "any serious debate" over the role of Islamic fundamentalism or radicalism in the region "is greatly impeded by the fact that very little concrete information is available. In the few instances where field research has been carried out, it has been based on relatively small samples" that are not representative of the underlying population and often produce contradictory results (100). Using the Pew Research Center's The World's Muslims dataset, I therefore examine whether either of these claims holds up to closer scrutiny.
The Pew dataset contains data based on 4,708 face-toface interviews conducted between November 24, 2011, and February 25, 2012, with self-identified Muslims in Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan. Lack of accessibility prevented interviews in Turkmenistan, and only seven ethnic Turkmen respondents appear in the data for the other four countries. I therefore limit my analysis to the 4,517 observations for ethnic Kazakhs, Kyrgyz, Tajiks, and Uzbeks. 1

Assessing Religiosity
There is no consensus regarding how best to measure religiosity in Islam (Abu-Raiya and Hill 2014). Survey-based studies use any number of questions, including frequency of prayer and whether one believes in Allah. Most studies, including those based in Central Asia, typically use just one question, making comparability across studies problematic. It is also extremely difficult to assess what scholars, policymakers, or average citizens precisely mean when they claim that Tajiks and Uzbeks are more "deeply religious." But this claim does seem to commonly encompass the notion that Tajiks and Uzbeks are more likely both to engage in a range of practices that typically signal piety-such as attending mosque and praying frequently, fasting, and giving zakat (alms, one of the five pillars of Islam)-and to reject "traditional" beliefs and practices such as pilgrimages to the shrines of saints or speaking to the souls of ancestors.
In an effort to capture as much of the breadth of this claim as possible, I examine religiosity using 16 separate questions found in the Pew survey. Table 1 lists each question, provides information about the possible responses to each question, and displays the range of values assigned to each. In all cases, higher values indicate greater scripturalist, or orthodox, religiosity. An affirmative answer to the last six questions, each of which addresses acceptance of traditionalist beliefs, signals a lower level of scripturalist religiosity. In other words, if respondents say that pilgrimages to shrines, devotional dancing, and so on are acceptable under Islam (answering "yes"), they are considered more traditionalist and less scripturally oriented (less "religious," in simplified terms). Full information on question wording and responses in the Pew surveys is available in the online appendix (found at http://www.rebe kahtromble.net/research/).

Assessing Radicalism
As with religiosity, there is no expert consensus regarding how to conceptualize radicalism (Borum 2011, 9;Pisiou 2013), and Central Asian leaders who speak of radicalism NOMADIC TRADITIONALISTS AND SEDENTARY SCRIPTURALISTS (or extremism) tend to use the term to effectively mean "any attempt by a religious figure or organization to participate in public life" (Khalid 2003, 588; see also Horsman 2005). Still, in both Central Asia and wider policy debates, uses of radicalism tend to "center around two different foci," identified by Tinka Veldhuis and Jørgen Staun (2009): "on violent radicalisation, where emphasis is put on the active pursuit or acceptance of the use of violence to attain the stated goal," and "on a broader sense of radicalisation, where emphasis is placed on the active pursuit or acceptance of far-reaching changes in society" (4). There is, however, a third dimension to radicalization that gets particular emphasis in Central Asia's ethno-religious narrative: a sense of rigidity and closed-mindedness that supposedly contradicts the moderation and general openness to ideas and beliefs purportedly inherent in the traditional approach to religion in the region.
To capture each of these three foci, I examine a series of 12 questions, four of which address religious openness, seven of which address far-reaching changes in society, including questions on the political system and punitive criminal practices, and one that speaks to support for violence against civilians to "defend" Islam. Table 2 lists each of the questions, as well as the range of answers and numerical values assigned to them. In all cases, higher values represent more "radical" views.
Due to the risk of repression in Uzbekistan, particularly sensitive political questions (marked in the table) were not asked in that country. Two additional questions (also marked in the table) had substantially different wording in Uzbekistan, and I therefore chose to exclude the Uzbekistani data for these questions. However, large numbers of ethnic Uzbeks served as respondents in Kyrgyzstan, making it possible to reach statistically significant conclusions regarding the impact of Uzbek ethnicity, even when lacking data from Uzbekistan. Again, full question wording and responses are available in the online appendix (found at http://www.rebekahtromble.net/ research/).
In adopting a multi-pronged conceptualization of radicalism, I do not mean to imply my own approval for using any of these questions, separately or in combination, as "true" measures of radicalism. A normative debate about what should be considered "radical" is beyond on the scope of this study. By examining each of these questions I simply seek to capture as much as possible of the meaning of the term as it is used within Central Asia.

Construct Validity
Some questions used in this analysis (regarding both religiosity and radicalism) may rightfully raise concerns regarding construct validity. For instance, the answer given to the question "How much does your life reflect the Hadith and Sunna?" (see Table 1) might, at least in part, reflect the respondent's degree of (im)modesty, rather than simply his/her level of religiosity. And responses to the question "How comfortable would you be if a daughter of yours married a Christian?" (see Table 2) might reflect respondents' comfort with a daughter marrying at all and not just their concern about marriage to a Christian per se. 2 Moreover, because the question "Do you favor or oppose making the sharia the law of the land in your country?" (Table 2) does not offer a definition of "sharia"-a highly contested termrespondents who support making sharia the law of the land may have a more "moderate" vision of term than those who oppose it. Readers should therefore approach the results related to these questions with a degree of skepticism. However, given the lack of consensus on how to define and measure either religiosity or radicalism, and because it is particularly important to understand how these different definitions and measurements might affect our understanding of the ethno-religious discourse in Central Asia, I have chosen to include a wide battery of relevant questions available in The World's Muslims survey.

Methods of Analysis
In order to test the two key claims discussed in this study, I subject the data for each of the religiosity questions listed in Table 1 and each of the radicalism questions listed in Table 2 to a series of one-way analysis of variance (ANOVA) tests to reveal whether there is a statistically significant difference in the mean scores for religiosity and radicalism among the four ethnic groups. I then scrutinize the factors that are associated with respondents' support for violence against civilians. Here I am interested in identifying whether ethnicity indeed plays a role in explaining support for violence once one controls for the impact of other factors commonly found in the literature on terrorism. I test the impact of ethnicity in two logistic regression models where the dependent variable is respondents' answer to the following: Some people think that suicide bombing and other forms of violence against civilian targets are justified in order to defend Islam from its enemies. Other people believe that, no matter what the reason, this kind of violence is never justified. Do you personally feel that this kind of violence is often justified to defend Islam, sometimes justified, rarely justified, or never justified? "Never" and "rarely" responses are coded 0; "sometimes" and "often" are coded 1. 3 Drawing on the empirical literature examining causes of support for terrorism (see Berger 2014;Fair et al. 2012;Tessler and Robbins 2007;Zhirkov et al. 2014), I include the following additional variables that control for the impact of religiosity, economic and political grievances, worldview, and socio-demographic factors: • Religiosity-assessed using three common measures found in the terrorism literature, each of which taps into a different dimension of personal religiosity (Fair et al 2012;Shapiro and Fair 2009): • Belief in Allah (0/1) • Importance of religion in one's life (0-3, "not at all important" to "very important") • Frequency of prayer (0-7, "never" to "all five salah daily") • Personal grievances • Personal economic situation (0-3, "very bad" to "very good") Note: † Question not asked in Uzbekistan. † † Uzbekistani data excluded due to difference in question wording.

NOMADIC TRADITIONALISTS AND SEDENTARY SCRIPTURALISTS
• Ladder of life-respondents asked to place their current situation on a "ladder of life," with "the best possible life" at the top (10 points) and "the worst possible life" at the bottom (0 points). • Future prospects-respondents asked where they see their prospects on this ladder in five years. Future prospects subtracts ladder of life response from this response. • Societal grievances • Economy (0-3, "very bad" to "very good") • Unemployment (0-3, "not a problem at all" to "very big problem") • Political corruption (0-3, "not a problem at all" to "very big problem") • Say in government-"People like me don't have any say about what the government does." (0-3, "completely disagree" to "completely agree") • Worldview • Urban (0/1, rural/urban)-proxies respondents' exposure to diverse peoples and perspectives. • Christian hostility-respondents asked how many Christians in their country they believe are hostile to Muslims (0-5, "none" to "all"), capturing feelings of fear and animosity toward religious "others." This question was not asked in Tajikistan "Don't know" or non-responses were treated as missing values for all of the statistical analyses. Table 3 presents the results of the ANOVA tests for level of religiosity. The mean response to each question is listed for all four ethnic groups, and superscripts indicate whether there is a statistically significant difference between the means for different groups. Table 4 provides a simplified visualization of these results, rank ordering the ethnic groups based on their mean responses to each question. Thus, in response to the question "Do you believe in God, Allah, and his prophet Muhammed?" both tables reveal that Tajiks and Uzbeks responded in the affirmative most often, followed by Kyrgyz and then Kazakhs. This finding is consistent with the first claim in the wider ethno-religious narrative-that Tajiks and Uzbeks are more religious than Kazakhs and Kyrgyz. Indeed, this claim is verified for all three baseline questions. Religion is most important to the lives of Tajik respondents, followed by Kyrgyz and Uzbeks, and then Kazakhs. Uzbeks said their lives most reflect the Hadith and Sunna, with Kyrgyz and Tajiks next, and Kazakhs again prove least religious. Thus, if we were to consider only these baseline measures of religiosity, it would be quite easy to conclude that the long-sedentary populations are indeed more religious than the historically nomadic groups in Central Asia.

FINDINGS Comparing Ethnic Groups
However, if we turn to either religious practices or opinions about the acceptability of traditional practices, the picture becomes more complex. Tajiks and Uzbeks do pray most often and are most likely to fast during Ramadan. But, on average, Uzbeks attend mosque and read the Koran least often, while Kyrgyz read the Koran most frequently and are the most likely to give zakat. Moreover, Kyrgyz were least likely to say that various traditional practices are acceptable under Islam, with the highest mean responses for four of the six applicable questions. In fact, Kyrgyz were about three times less likely to regard pilgrimages to the shrines of Muslim saints as acceptable than were members of any of the other ethnic groups. On the other hand, across all questions, Kazakhs proved to be least religious in all but one case (frequency of mosque attendance), though in a number of instances (pilgrimage to Mecca; frequency of reading the Koran; acceptability of pilgrimages to shrines, devotional dancing, and reciting poetry or singing in praise of Allah), this last-order ranking is shared with Tajiks, Uzbeks, or both.
Moving to the second core claim within Central Asia's ethno-religious discourse, Tables 5 and 6 present the results of the ANOVA tests for degree of radicalism.
Once again, one might draw differing conclusions based on the type of questions examined. If considering the first five questions regarding religious rigidity, the claim that Tajiks and Uzbeks are less moderate stands. In this survey, Tajiks and Uzbeks were less comfortable with a daughter marrying a Christian. They were also more likely to say that Islam is the only path to heaven and that there is only one true way to interpret the teachings of Islam. When asked whether the sharia should be open to multiple interpretations, Tajiks were most likely to respond that no, there is only one true interpretation, while Kyrgyz and Uzbeks were next most likely. For each of these five questions, Kazakhs were least likely to give responses that pointed toward religious rigidity. 362 TROMBLE However, if we turn to the remaining questions-those regarding the role of Islam in politics, punitive measures for "improper" or criminal behavior, and support for violence against civilians in the name of Islam-the conclusions are not so clear cut. Though Kazakhs again gave the least radical responses to each of these questions, on average, the responses of ethnic Kyrgyz place them as the most radical among the four ethnic groups for six of these eight questions, including support for the use of violence against civilians. It is important to note that very few Kyrgyz actually supported such violence. On a scale of 0-3, with zero indicating that violence against civilians is never justified and three indicating that violence is often justified, the mean score for Kyrgyz was just 0.41. Still, this is 1.6 times greater than the mean for Uzbeks, 2.9 times greater than the mean for Tajiks, and 5.1 times greater than the mean for Kazakhs.

Explaining Support for Violence
This surprising finding deserves further investigation. Table 7 presents the results of two logistic regression models that examine whether ethnicity indeed impacts support for violence in the name of Islam once one introduces a number of controls. Table 8 shows the predicted probability of responding that violence is sometimes or often justified for each of the statistically significant independent variables in the model.
At aggregated levels, only the variable for Kyrgyzstani respondents has a statistically significant association with support for violence. The differences among various ethnic groups-no matter where the respondents reside-show no such association. Indeed, even in Kyrgyzstan, the only country that had substantial numbers of ethnic minority respondents, the interaction term for Kyrgyzstan and Uzbek reveals that there is no difference between these two ethnic groups within the country. 4 Instead, the model suggests that certain individual-level variables-belief in Allah, prayer frequency, urbanity, the perception of one's personal future prospects, as well as the perceived level of Christian hostility in a country-are correlated with support for violence. 5 Though two of the three religiosity measures are associated with support for violence, dynamics are not as predicted by the traditionalist/ scripturalist narrative. While greater piety in the form of more frequent prayer is associated with greater support for violence, the overall impact is negligible. A shift from never praying to praying all five salah results in just a 3 percent greater probability of supporting violence. Belief in Allah and the Prophet, on the other hand, actually lowers support for violence, with Kyrgyz believers in Model 1 showing a 20.4 percent and

NOMADIC TRADITIONALISTS AND SEDENTARY SCRIPTURALISTS
Uzbek believers in Model 2 a 1.4 percent lower probability of supporting violence. Thus, while we cannot wholly dismiss the link between religiosity and extremism, one cannot simply conclude that greater religiosity is inherently and substantially tied to radical opinions. Other factors impacting support for violence fall within the categories of (a) political and economic grievances and (b) openness and exposure to "others." In the former instance, the model shows that those who hold little hope for their future prospects are much more likely to support violence. This is consistent with previous research, which has found that the perception of a stall or backslide in life conditions can generate frustration that promotes radicalization (Shapiro and Fair 2009). However, the statistical models suggest that, overall, the most significant impact on one's support for violence is linked to an individual's exposure to and view of outsiders or "others." Support for violence is 7.7 percent more likely among rural respondents in Model 1 and 6.1 percent more likely in Model 2, and it is 50.8 percent more likely among those who believe all Christians in their country are hostile toward Muslims compared to those who perceive no such hostility.
In sum, then, the data show no association between ethnicity and support for violence and suggest that if we want to better understand the antecedents of radical beliefs, it is more appropriate to concentrate on the impact of certain individual-rather than group-level variables.

Data Limitations
Of course, the data used in each of the above statistical analyses carry limitations. For one, it is impossible to determine the precise reasons for the country-level differences between Kyrgyzstan and its neighbors found in the two regression models. Is the greater level of support for violence in Kyrgyzstan a result of political, economic, or cultural dynamics? Unfortunately, a single country-level dummy variable provides no guidance. Future work with longitudinal data is needed to permit more precise analysis of system-level variation.
The need for longitudinal analysis points to another limitation of these data. Because they offer a single snapshot in time, these data cannot speak to whether the region's ethno-religious dynamics have changed over time nor to how religiosity-or any other factors, for that mattermay have impacted support for violence in previous eras. As such, I intentionally limit the extent of my claims to recent trends. While the essentialism inherent in the longstanding ethno-religious narrative should itself give pause, the data presented here provide direct evidence of its limitations only for recent developments in the post-Soviet period of Islamic revival.
A final limitation arises from the potential impact of social desirability effects and fear of repression, which are likely to bias the data downward-that is, toward expressions that suggest less religiosity and radicalism than may be accurate (Drakos and Gofas 2006). Though scholars have begun to use survey data to explore questions related to religiosity and political Islam in Central Asia (Collins and Owen 2012;Shaykhutdinov and Achilov 2014), no research has explored social desirability and the effects of fear in the region. However, recent work in Russia suggests the problem could be quite severe. Kirill Kalinin (2016), for instance, finds that electoral support for Vladimir Putin in 2012 was inflated in public opinion surveys by as much as 15 percent, and the Levada Center found that roughly onequarter of Russians were afraid to express their opinions to pollsters. 6 And yet the variations in responses observed among Central Asia's ethnic groups in this survey are not clearly and consistently tied to the varying levels of repression in the region. Table 9 offers several measures of repression in the four countries examined here, from general measures of political freedom to specific concerns about religious repression. The Freedom House scores for political freedom and civil liberties are drawn from 2011 and 2012, the period during which The World's Muslims survey was conducted. Based on these measures, Uzbekistan is the most authoritarian state and Kyrgyzstan the least, though the latter is relatively close to both Kazakhstan and Tajikistan. However, within Kyrgyzstan there is a marked difference between the level of repression faced by the titular Kyrgyz population and ethnic Uzbeks. For many years "ethnic Uzbeks living in Kyrgyzstan [have been] under constant surveillance by the authorities…. People who begin to practice Islam in their daily lives are immediately treated suspiciously, can be arbitrarily arrested, and many have been imprisoned for 3-5 years," (IHFHR and Memorial 2006, 6). These tactics only intensified following violent clashes between Kyrgyz and Uzbeks in 2010 (Human Rights Watch 2012).
The data on religious freedom provided by the Association of Religious Data Archives (ARDA) represent average scores derived by coding the U.S. State Department's International Religious Freedom Reports from 2003, and 2008. Based on these data one might conclude that Kazakhstan represses religious activity much more severely than Tajikistan. However, the ARDA data miss key developments between 2008 and 2012. In October 2011, Kazakhstan increased the registration requirements for religious groups and began shutting down small religious communities (Human Rights Watch 2013, 451). In Tajikistan new restrictions were even more dramatic. In 2009, the country introduced a law that banned all religious activities by unregistered groups; limited the number, type, and location of mosques permitted throughout the country; required state approval of all religious materials; and severely restricted institutions of Islamic education (Human Rights Watch 2011, 476-77). In 2011, Tajikistan further banned study at foreign religious institutions unless approved by the state, and prohibited anyone under the age of 18 from taking part in almost all religious activities. Each of these restrictions was accompanied by increased monitoring of mosques and prayer houses and more intense harassment of people wearing religious clothing (Human Rights Watch 2012). Taken together, these developments suggest that religious repression was just as severe (if not more so) toward Muslims in Tajikistan as in Kazakhstan at the time of the Pew survey.
Overall, then, we might conclude that residents of Uzbekistan faced the greatest fear of repression and Kyrgyzstani residents the least, with the qualification that Uzbek responses come from residents of Kyrgyzstan (as they were not asked in Uzbekistan)-it is notable that Uzbeks give more moderate responses than Kyrgyz in just three of seven instances, and the two groups are statistically indistinguishable from one another in all other cases. Moreover, as the ANOVA results show, Tajiks' responses to two of these extremely sensitive questions were substantially more radical than those of Kyrgyz; two were significantly less so, and the remainder were statistically indistinguishable from Kyrgyz. In short, then, there is no clear correlation between question responses and level of repression.
To be sure, without research into the direct impact of social desirability and fear of repression on survey results across Central Asia, one cannot draw any firm conclusions regarding their influence on these data. However, fear of repression alone does not seem to provide a straightforward explanation of the results found here.

CONCLUSION
While additional evidence is indeed needed to address data limitations and to explore the ethnicity-religiosity-radicalism nexus further, the Pew Research Center data are by far the most expansive and representative data available to date, and as such, their results deserve serious consideration. Whether the ethno-religious narrative holds up to empirical scrutiny largely depends on how one conceptualizes and assesses both religiosity and radicalism. In some instances Tajiks and Uzbeks do prove more religious and less moderate than their historically nomadic neighbors. They are most likely to believe in Allah and the Prophet, feel religion is important, and regard their lives as, to at least some extent, reflecting the Sunna and Hadith. Thus, if religiosity is understood and measured exclusively on the basis of selfreflexive understandings of the role of religion in one's life, the narrative does hold up to scrutiny. But our examination should not end there. After all, scholars, commentators, and policymakers have long used religious practices as a benchmark of religiosity. Indeed, Central Asian officials use outward signals of religious piety to determine the "risk" posed by individuals to the carefully managed balance between religion and state power. Moreover, the notion that historically sedentary and nomadic populations are separated by the degree to which they accept or reject traditional practices dating from pre-Islamic times is central to the ethnoreligious narrative. And yet if we assess religiosity using either of these measures, the data do not support the claim that Tajiks and Uzbeks are naturally more religious than other Muslims in the region. While this claim does hold for the comparison of Tajiks and Uzbeks to Kazakhs, it does not hold for the comparison with Kyrgyz, casting at least some doubt on the broader narrative regarding the impact of nomadism on contemporary religious life.
The same dynamics play out when looking at the data on radicalism. If we define radicalism exclusively as rigidity of faith, then Tajiks and Uzbeks could be called more radical overall. But such a narrow definition seems inappropriate. While religious rigidity plays a key role in Central Asia's ethno-religious narrative insofar as scripturalists are distinguished from traditionalists by the former's supposed rigidity and the latter's purported openness, this is certainly not the core of radicalism itself, despite popular rhetoric to the contrary. As hotly contested as the term may be, few experts would contend that rigidity of beliefs constitutes radicalism. In other words, to some, rigidity may be necessary, but is certainly not a sufficient condition for radicalism. If instead we look to the desire for changes to the political system, support for harshly punitive measures for social and criminal wrongdoing, and, most important, support for violence against civilians in the name of Islam, the claim that Tajiks and Uzbeks are most likely to radicalize does not hold up to scrutiny. In fact, when support for violence is subjected to further analysis, the results show that ethnic differences disappear with the introduction of country-and individual-level controls. And though one's religiosity does impact support for violence, crucially, this impact is not as straightforward as is typically implied by the traditional/ scripturalist narrative. Those who pray more frequently do show slightly more support for violence in the name of Islam, but those who profess basic belief in Allah and the Prophet Muhammad also show much less support for such violence overall. These results are consistent with the larger literature on extremism and terrorism, which finds that, despite the widespread, popular discourse linking religiosity and radicalism worldwide, "religiosity writ large is a poor predictor" of support for violence in the name of Islam (Shapiro and Fair 2009, 102). Instead, socio-economic and political grievances-which generate disillusionment with and embitterment toward the state-as well as factors related to the openness or insularity of one's worldview, provide better predictors of support for violence.
Indeed, these latter variables may help explain some of the differences found in the ANOVA comparisons between Central Asian ethnic groups. As Eric McGlinchey (2009) observes, of the four countries examined in this study, Kyrgyzstan, Tajikistan, and Uzbekistan "are still struggling to meet the basic needs of their populations" (27). And where state failures are greatest, a turn toward faith helps people fill both practical and socio-psychological needs (McGlinchey 2009;Montgomery 2015). Thus, the fact that Kazakhs fairly consistently demonstrate the lowest religiosity and radicalism scores may have less to do with intrinsic ethnic qualities and more to do with the relative socio-economic security in which they live.
What is more, the Kazakh population is the most urbanized across the region. Just over half of Kazakhstani residents live in urban environments, compared to between roughly one-quarter and one-third of the populations in the remaining three countries. Though the goal of this study was to assess the impact of ethnicity on religiosity and radicalism in Central Asia-not to establish clear alternative causes-and much more extensive  Note: † Kazakhstan serves as the reference category. † † Tajik serves as reference category.*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.00 NOMADIC TRADITIONALISTS AND SEDENTARY SCRIPTURALISTS research is needed on this matter, based on the results of the initial regression analyses offered here, one might postulate that the relative insularity generated by rural life in many pockets of the region, rather than ethnicity per se, may contribute to more parochial attitudes toward "outsiders" or "others" and to rigidity of opinions concerning religious and political life.
Ultimately, if there was, as other scholars have argued, already prima facie reason to be skeptical of simplified versions of the Central Asian ethno-religious narrative, the empirical evidence presented in this study should enhance that skepticism-at least for the most recent period of post-Soviet development. When we are dealing with such a powerful narrative-one that helps justify repression and may promote tension and suspicion among neighbors-it is worth reconsidering whether historical differences between Central Asia's sedentary and nomadic populations continue to generate a clear fault line that defines religious and political dynamics in the region.

NOTES
1. The remaining 184 interviews were with members of a number of different ethnic groups, including Azeris, Dungans, Kurds, Iranians, Russians, and Tatars. The region's ethno-religious narrative has not extended to cover these groups, and were it extensible, there are too few observations for any single ethnicity to draw statistically significant conclusions. See the original Pew documentation for further details regarding sampling techniques (www.pewresearch.org). 2. I am grateful to the article's first anonymous reviewer for these insights. 3. Factor analysis revealed this to be the most appropriate grouping of responses; treating the responses as three or four ordered categories violated the proportional odds assumption. However, ordered logit models produced very similar outcomes to the logit models used here. 4. Since there are only five respondents of Kazakh and one respondent of Tajik ethnicity from Kyrgyzstan, this variable effectively measures the difference between Uzbek and Kyrgyz respondents within Kyrgyzstan. 5. Say in government does have a statistically significant impact on support for violence. However, as Table 8 demonstrates, the relationship is neither linear nor consistent across models. 6. See http://www.kommersant.ru/doc/2897562 (accessed September 4, 2016).