The demographic literature on union formation in post-communist Europe typically documents retreat from marriage and increase in cohabitation. However, sociological and anthropological studies of post-Soviet Central Asia often point to a resurgence of various traditional norms and practices, including those surrounding marriage, that were suppressed under Soviet rule. We engage these two perspectives on union formation by analyzing transition to first marriage in Kyrgyzstan both before and after the collapse of the USSR. We use uniquely detailed marriage histories from a nationally representative survey conducted in the period 2011–2012 to examine the dynamics of traditional marital practices among that country’s two main ethnic groups—Kyrgyz and Uzbeks—focusing on trends in arranged marriages and in marriages involving bride kidnapping. The analysis reveals instructive ethnic and period differences but also indicates an overall decline in the risks of both types of traditional marriage practices in the post-Soviet era. In fact, although the decline has characterized all marriage types, it was more substantial for traditional marriages. We interpret these trends as evidence of continuing modernization of nuptiality behavior in the region.
Although changes in nuptiality regimes and union formation take different shapes in different societies, for many settings—including those of the former USSR—the demographic literature documents decline of marriage rates and increase in nonmarital cohabitation in recent decades (e.g., Gerber and Berman 2010; Katus et al. 2008; Kennedy and Bumpass 2008; Kierman 2001). In societies with traditionally high prevalence of arranged marriages—that is, marriages in which the decision to marry is not made by the groom and the bride—studies point to declining involvement of parents in marriage decisions and a rising share of marriages based on the groom’s and the bride’s choice (Bahramitash and Kazemipour 2006; Ghimire et al. 2006; Heaton et al. 2001; Thornton et al. 1994; Zang 2008). These fundamental shifts in nuptiality regimes are said to be both products and engines of broader social and demographic changes, such as the rising status of women and declining fertility (Feyisetan and Akinrinola 1991; Ghimire and Axinn 2013).
One part of the world that does not seem to fit neatly into this developmental paradigm is post-Soviet Central Asia, a predominantly Muslim region comprising five nations with a total population of some 65 million people. The most prominent demographic feature of Central Asia is its comparatively high fertility; it lagged behind most of the Soviet Union and Eastern Europe in demographic transition and met the breakup of the USSR with fertility well above replacement level. The sociological and anthropological literature on this region has often pointed to retraditionalization of marital customs and practices after the collapse of the Soviet system in the early 1990s. Specifically, studies have noted the resurgence of polygamy and increase in early marriage, marriage among close relatives, religious marriage, traditional wedding ceremonies, and bride payments throughout the region (e.g., Dadabaev 2007; Handrahan 2001; Koroteyeva and Makarova 1998; Roche and Hohmann 2011; Werner 1997). The scholarship on Kyrgyzstan, the focus of our study, has also depicted a major revival of bride kidnappings (e.g., Amsler and Kleinbach 1999; Kleinbach 2003). At the same time, several recent studies have produced evidence of a demographic modernization of Central Asia, such as an increase in nonmarital fertility, spread of cohabitation, marriage postponement, and retreat from marriage (e.g., Agadjanian et al. 2013; Denisenko 2004; Denisenko and Kalmykova 2011; Dommaraju and Agadjanian 2008; Nedoluzhko 2011)—the trends that to a greater or lesser extent have characterized other, non-Muslim parts of the former Communist world.
This changing nuptiality behavior is part of the general societal and cultural transition propelled by the breakup of the USSR and the transformation of the former Soviet republics into independent nations. Thus, the literature on Central Asia has devoted considerable attention to what is often characterized as a retreat from the modernizationist policies of the Soviet era and a revival of pre-Soviet traditions and norms, especially those rooted in the Islamic religion, and a concomitant retraditionalization of many aspects of societal life (e.g., Brusina 2008; Dadabaev 2007; Koroteyeva and Makarova 1998; Louw 2013; Phillips and James 2001; Tazmini 2001). At the same time, it has also been argued that the breakup of the Soviet Union has led to greater openness of the region to Western cultural and social influences (e.g., Agadjanian and Dommaraju 2011). These parallel, even if conflicting, processes have resulted in a unique mix of modernity and tradition in the region. Apart from this complex and fluid sociocultural milieu, the volatile economic situation following the breakup of the USSR has also played a substantial role in marriage dynamics. The demographic literature on former Communist countries has typically entertained two perspectives on the role of economic factors. The first perspective considers the relative importance of the early post-Communist economic crisis and, in some cases, of subsequent economic recovery and stabilization. It generally fits with the pro-cyclical view on the relationship between marriage dynamics and economic fluctuations (e.g., Agadjanian et al. 2013; Heleniak 1995; Philipov and Jasilioniene 2008). The second perspective suggests that the transition to the market economy has generated economic circumstances that paved the way to the spread of norms, values, and attitudes around family formation and childbearing that had prevailed in Western societies and are conducive to marriage decline (e.g., Frejka 2008; Sobotka et al. 2003). A similar interpretation views economic crisis at the onset of the post-Communist era as galvanizing the demographic change that had already been underway prior to the collapse of communism (Gerber and Berman 2010; Hoem et al. 2009).
In addition to sociocultural and economic forces, another possible factor in marriage dynamics is the legislative chaos of the transitional period that could have given an impetus to the resurge of marital customs and practices that were banned under Soviet rule. Importantly, the sociocultural, economic, and legislative dynamics in the post-Soviet era have been organically interconnected; although we relate our results to the multifaceted nature of post-Soviet societal change, we are unable to fully disentangle the interwoven effects of cultural, economic, and other factors.
Engaging the modernization versus retraditionalization debate, our study seeks to examine how the balance between different marriage types has changed in Kyrgyzstan, a Central Asian nation of some 5.5 million inhabitants with a gross national income per capita of $1,210 USD (World Bank 2013) since it became independent from the Soviet Union in 1991. The study covers the late Soviet period as well as two post-Soviet decades; the first post-Soviet decade, particularly its early part, was marked by radical sociopolitical transformations and dramatic economic decline, and the following decade saw a gradual social stabilization and economic recovery and growth. Using unique recent survey data, we analyze marriage dynamics by estimating competing risks of arranged versus nonarranged (choice) marriage among Kyrgyzstan’s two largest Muslim ethnic groups: Kyrgyz, the nation’s titular ethnicity; and Uzbeks, its biggest minority. For ethnic Kyrgyz, which is a historically nomadic group, we also examine the dynamics of marriage involving bride kidnapping—a traditional, pre-Islamic practice common among nomads in Central Asia and elsewhere that can be either forced or consensual.
Conceptualization and Hypotheses
In the rich cross-national literature on nuptiality, Central Asia has received little attention. Research on union formation in this part of the post-Soviet world is scant and often limited to examinations of aggregate marriage trends (e.g., Denisenko 2004; Denisenko and Kalmykova 2011). Earlier studies have been generally limited to contrasting natives and Europeans (i.e., descendants of voluntary and forced migrants mainly from Russia), two population segments with vastly different cultural and demographic backgrounds (e.g., Agadjanian 1999; Agadjanian and Makarova 2003; Agadjanian et al. 2008; Nedoluzhko and Agadjanian 2010). Only one recent study has looked at ethnic-specific probabilities of first marriage among indigenous groups in Kyrgyzstan and Kazakhstan (Agadjanian et al. 2013). That study showed an overall decline in first-marriage probabilities in the decade-and-a-half after the dissolution of the Soviet Union, but because of the lack of appropriate data, the authors could not investigate trends in different types of marriage.
Our study aims to contribute to the literature on nuptiality of autochthonous Central Asian ethnic groups in general, and on the dynamics of traditional marriages in particular. We compare trends in arranged and nonarranged marriages among the two largest native ethnic groups in Kyrgyzstan—Kyrgyz and Uzbeks—which, according to the 2009 national population census, accounted for 70.9 % and 14.3 % of the country’s population, respectively (Mkrtchyan and Sarygulov 2011:85). We define as “arranged” those marriages in which at least one of the spouses (mostly the wife) did not take part in the marriage decision. For ethnic Kyrgyz, we also analyze trends in marriages involving bride kidnapping. We differentiate between marriages with forced kidnapping, which by definition are a specific type of arranged marriage, and marriages with consensual, or mock, kidnapping (elopement). We contrast kidnapping-based marriages with marriages that did not involve bride kidnapping. Nonarranged marriages and marriages without bride kidnapping are defined in our study as “choice” marriages: that is, marriages in which both the bride and the groom participated in the marriage decision.1 We refer to arranged marriages and marriages involving forced bride kidnapping as “traditional” marriage types. In our conceptualization, mock kidnapping represents an intermediate marriage type that combines features of a traditional marital practice with those of choice marriage.
Although no demographic analyses of arranged marriages in Central Asia have been attempted, considerable sociological and anthropological literature has focused on emotional, ethical, social, and gender implications of bride kidnapping among the region’s historically nomadic groups: Kazakhs and, especially, Kyrgyz (e.g., Amsler and Kleinbach 1999; Borbieva 2012; Brusina 2008; Handrahan 2004; Kleinbach and Salimjanova 2007; Werner 2009). There have been attempts to estimate the incidence of bride kidnapping in independent Kyrgyzstan, yet these attempts are not based on representative samples (e.g., Amsler and Kleinbach 1999; Kleinbach 2003; Kleinbach et al. 2005).
Our hypotheses regarding temporal trends in traditional marriage address the modernization versus retraditionalization dilemma. Thus, under the modernization assumption, the incidence of traditional marriage should decline. On the contrary, the retraditionalization scenario would imply an increase in the incidence of traditional marriage. At the same time, from the modernization perspective, the dynamics of mock kidnapping could differ from those of arranged marriage and marriage involving forced kidnapping. If marriage with mock kidnapping is a transitional form between marriage based on forced kidnapping and choice marriage, then under the modernization scenario, the incidence of marriages involving mock kidnapping should first increase and then decrease.
Additionally, changes in the dynamics of traditional marriages could be group-specific: groups that are culturally more conservative could display greater adherence to arranged marriages than culturally less-conservative groups. Here our hypotheses are guided by the assumption that Uzbeks, a traditionally sedentary agricultural people with a longer and stronger influence of Islam, are in general more culturally and demographically conservative in comparison with Kyrgyz, traditional nomads, whose sedentarization and Islamization are historically recent. The adherence to traditional practices among Uzbeks may also have been enhanced by their minority status in a country where the titular group is increasingly favored. Thus, several studies have suggested that ethnic or other minorities in settings where they feel marginalized or threatened often tend to adhere to their traditions, including those related to marriage and reproduction, as a means of preserving their group identity and cohesiveness, and eventually their numerical strength (e.g., Anson and Meir 1996; Fargues 2000). We therefore expect that Uzbeks, ceteris paribus, will be more inclined to retraditionalization of marital practices. Accordingly, we hypothesize that the incidence of arranged marriage for the two ethnic groups in our study will change at different paces. Specifically, under the retraditionalization assumption, Uzbeks should exhibit a stronger revival of arranged marriages whereas, under the modernization scenario, they should have a slower decline in arranged marriages compared with Kyrgyz.
The anticipated trends in different marriage types may not manifest themselves uniformly throughout the entire post-Soviet period. Thus, it is possible that reemergence of traditional marriage forms would be particularly pronounced in the early years after the collapse of the Soviet Union, when the decades-long restrictions on traditional norms and practices, of both Islamic and pre-Islamic nature, were swiftly removed. After this initial traditional revival, marriage practices could be expected to settle back into a modernizationist path. Also, divergent trends in arranged versus choice marriages in the early post-Soviet period may reflect the impact of the economic crisis that characterized it. Thus, although the literature typically suggests a decline of marriage during hard economic times (e.g., Frejka 2008; Galloway 1988; Perelli-Harris 2008; Weir 1984), bride kidnapping could have been fostered by the post-Soviet crisis because kidnapping, which is usually followed by a quick and simplified wedding with a reduced bride price, might help to lower the financial burden of marriage (see, e.g., Handrahan 2004; Kleinbach and Salimjanova 2007; Lockwood 1974; Werner 1997). Extending this logic into the period of post-crisis economic recovery, one could argue that with an improvement in families’ economic fortunes, more men would be able to afford bridal payments, and therefore, the incidence of bride kidnapping should decline as the economy stabilizes and resumes growth.
Our data come from the nationally representative household survey Socio-Economic and Migration Processes in Kyrgyzstan conducted in late 2011 and early 2012. The survey interviewed men and women aged 18–49, with one respondent per household (N = 2,032). The survey data set contains respondents’ marital, childbearing, migration, employment, and educational histories as well as various other characteristics of the respondents, of their parents and spouses/partners (if any), and of their households at the time of interview. To estimate risks of marriage, we employ the uniquely rich information from respondents’ marital histories. To our knowledge, the survey that supplies our data is the first nationally representative sample survey that distinguished between arranged marriages and choice marriages in Central Asia. It is also a pioneering survey of this kind in assessing the incidence of forced and mock bride kidnapping.
To account for the effects of individual socioeconomic characteristics and place of residence, we use information on the respondents’ employment, educational, and migration experiences. We restrict the analysis to two autochthonous Central Asian ethnicities: Kyrgyz and Uzbeks. The survey sample also included Russians and other respondents of European roots; however, these respondents are not part of our analysis as they are very unlikely to experience arranged or forced marriages. We also exclude from the analysis representatives of other ethnic groups whose numbers in the sample are too small to allow for sound comparisons.
Several limitations of our data must be acknowledged. Most importantly, detailed information on union type—that is, arranged, forced, or choice marriage—is available only for current unions. We do not have such information for previous unions (if any) and therefore exclude from the analysis those respondents who were in a union at the time of survey but who had an earlier official or religious marriage as well as those who were not in a union at the time of the interview but had been married before (in total, 13.1 % of Kyrgyz and 7.7 % of Uzbeks in our sample). To test whether the exclusion of these respondents from the analysis affected our results, we estimated alternative sets of models under different imputation assumptions (including the extreme assumptions that all dissolved unions were arranged or all were choice marriages) and compared the results with the results of the models limited to respondents with no history of dissolved marriages. These comparisons revealed no substantial discrepancies between the results of the models presented here and those of the alternative models.
Other limitations concern the covariates used in the analysis. Thus, the survey did not collect data on socioeconomic status of respondents’ families of origin; we use “father’s occupation” as a proxy for family socioeconomic background. Also, the survey collected information only on respondents’ five most recent occupations, including occupation at the time of interview); therefore, employment histories of respondents who changed more than five occupations are truncated. However, the share of respondents for whom five occupations were recorded and thus who potentially could have richer employment histories is very small, constituting only 2 % of the total sample and 3 % of ever-employed respondents. Finally, the survey accounts for years, but not months, of start and completion of respondents’ various educational stages. In constructing the “educational enrollment” covariate, we assume that enrollment followed a standard academic year calendar. The timing of other events in respondents’ lives is defined with an accuracy of a month.
We employ event-history analysis to model risks of entry into different types of marriage. This method allows us to examine the risks dynamically and to account for the effects of time-varying covariates. We fit multiplicative intensity regression (or proportional hazard) models with a piecewise-constant baseline hazard (Hoem 1971, 1976). In addition, we model competing transitions to different types of marriage jointly and compare the effects of various factors across these transitions using an extension of this approach described in Hoem and Kostova (2008). All models are fitted in the STATA software.
The dependent variable in the analysis is the occurrence of first marriage, which includes both officially registered marriages and marriages formed through a religious ceremony but without official registration: the latter category constitutes only 3 % of all current marriages. Respondent’s age—or, more precisely, the number of months elapsed since his/her 16th birthday—is the basic process time variable. We define five 3-year age groups from 16 to 30 and a broader age group of those older than 30. We lump together all ages above 30 because in Kyrgyzstan, as in many similar settings, first marital unions are rarely formed after age 30.
Guided by our hypotheses on the trends in marriage formation, we divide the time of observation into five periods: 1980–1990, the decade preceding the disintegration of the USSR; 1991–1995, the period of particular severe societal shocks associated with the breakup of the Soviet empire; 1996–2000, when the economic decline continued but its magnitude became less dramatic; 2001–2005, a period of initial post-crisis stabilization and slow economic growth; and 2006–2011, a period when this growth somewhat accelerated (despite the global economic recession). Given the age range of the survey sample, the first period does not include ages above 28, the second period does not include ages above 33, and so on. Only the last period includes the entire age range of up to age 49. Because most first marriages take place at young ages, the uneven distribution of older respondents across the time periods should not be a problem.2
To account for ethnic differences in the transition to arranged versus nonarranged (choice) marriage, the corresponding models are standardized for respondents’ ethnicity: Kyrgyz or Uzbek. Other control variables are respondent’s gender, employment, educational enrollment and attainment, and type of place of residence. Educational attainment and enrollment of respondents’ as well as their employment status and place of residence are time-varying covariates. The lowest educational level in our analysis is basic secondary education; this category also includes several dropouts from secondary school. For the employment status of respondents, we define two levels: non-employed and employed. The place-of-residence covariate has three levels: the capital city (Bishkek), other urban, and rural.3 This covariate accounts for all relocations of respondents within the country; observations are censored at the time of first migration abroad. We exclude from the analysis 17 cases of respondents who moved into Kyrgyzstan from the neighboring countries of Uzbekistan and Tajikistan after age 16.
Finally, the socioeconomic status of respondent’s family of origin is approximated by the occupation of his/her father or stepfather when the respondent was about 15 years old. This time-fixed covariate differentiates among three occupational groups: higher status occupations (e.g., managers and highly qualified specialists), other nonagricultural occupations, and agricultural occupations. It also includes a “nonworking, not specified” category that constitutes about 12 % of the sample and is composed of respondents whose fathers were not working or were unemployed, respondents who did not know their fathers’ occupation, and respondents whose fathers were not alive or not around when they were 15 years old. The distribution of occurrences and exposure months by each covariate used in the analysis is presented in the appendix.
Table 1 presents the distribution of respondents who were married at the time of interview by ethnicity, gender, and marriage cohort according to marriage type. The table shows that more than 30 % of marriages in our sample can be classified as arranged. The share of such marriages is noticeably higher among Uzbeks, befitting the notion of greater traditionalism of that group. Among Kyrgyz, one-third of marriages involved bride kidnapping; in one-half of those marriages, kidnapping was forced. Period comparisons, the primary interest of our study, suggest unstable dynamics of traditional (arranged and kidnapping-based) marriages during the post-Soviet period. Yet, in both ethnic groups, the share of arranged marriages was the lowest in the last marriage cohort, and so was the share of marriages involving forced bride kidnapping among Kyrgyz.
Table 1 also depicts ethnic differences in marriage timing; in general, Uzbeks have lower ages at marriage compared with Kyrgyz. Both ethnic groups exhibit a tendency toward marriage postponement, which is particularly pronounced among men. Moreover, Table 1 suggests gender differences in reporting the involvement of grooms and brides in marriage decisions: men, especially among Uzbeks, generally appear to be less likely to admit that decisions about their marriages did not involve their input or that of their spouses (recall that “arranged” marriages includes marriages formed without the consent of the respondent or his/her partner, as reported in the survey). Kyrgyz women also more frequently reported bride kidnappings and were more likely to characterize kidnapping-based marriages as forced, compared with Kyrgyz men. Whereas these gender differences may reflect broader gender ideologies, they can also be explained, at least partially, by unbalanced distributions of men and women across various individual characteristics in our sample, which we control for in the following analysis.
Multivariate Results: Arranged Versus Nonarranged Marriages
Table 2 displays the results of three multivariate models, presented as relative risks. Model 1 combines transitions to arranged and nonarranged (choice) marriages: that is, the corresponding estimates refer to marriage in general, irrespective of its type. Models 2 and 3 are competing risks models, estimated separately. The results of all three models uniformly suggest that the risks of both arranged and nonarranged marriage declined considerably throughout the 2000s. At the same time, there was no substantial or statistically significant difference in marriage risks across the first three time periods defined in our study—the late Soviet years and the two 5-year periods of the first post-Soviet decade. It is also noteworthy that the decline in marriage risks was most pronounced in the case of arranged marriages compared with nonarranged marriages. Overall, these trends do not support the retraditionalization argument.
The results in Table 2 are also instructive with respect to ethnic differences. Although no gap between Kyrgyz and Uzbeks is noticeable in the risks of transition to choice marriage, the results for arranged marriage (Model 3) show a strong ethnic contrast, with Uzbeks having almost 50 % higher marriage risks compared with Kyrgyz. All three models in Table 2 uniformly suggest that women have higher risks of marriage compared with men, reflecting the fact that women enter marriage at younger ages than do men.
The effects of other covariates are also noteworthy. The effect of employment status is not significant in any of the models. Supporting the common finding that schooling conflicts with family responsibilities (e.g., Hoem 1986; Santow and Bracher 1994; Thornton et al. 1995), our results indicate considerably lower marriage risks among respondents currently in an educational institution compared with those respondents who had finished their education. This result is consistent in all three models. The estimates for educational attainment also uniformly suggest a higher propensity to marry among more-educated respondents. The positive association between education and marriage, as argues the literature on other settings (see, e.g., Oppenheimer 2003; Thornton et al. 1995), may indicate that educational attainment, as a marker of long-term socioeconomic prospects and of earning capacity, increases the value of a potential spouse on the marriage market.4 The risks of nonarranged marriage are comparatively high in rural areas; at the same time, the risks of arranged marriage seem somewhat higher in cities. The latter result appears counterintuitive, but the corresponding coefficient is not statistically significant precluding any inferences. Father’s occupation has an effect on arranged but not on nonarranged marriage. The highest risks of arranged marriage are among respondents whose fathers worked in agriculture, which probably reflects the effects of the traditional cultural background of a respondent’s family of origin.
To provide a more refined insight into period effects and possible ethnic variations in these effects, we estimate joint models of arranged and nonarranged marriage. The joint modeling allows for direct comparison of risks across competing transitions. The essence of this analytical technique is the formal inclusion of the decrement as an extra factor in the model, a factor that can be interacted with any of the predictor or control variables (see Hoem and Kostova 2008).
We start the presentation of the results of joint modeling with an interaction between marriage type and ethnicity (Table 3). These and any other results of joint modeling should be interpreted as averages standardized for the other covariates. The results for ethnic differences essentially confirm those presented in Table 2: even though there is no substantial difference in the risks of nonarranged marriage between the two ethnic groups considered in our analysis, Uzbeks have a 61 % higher average risk of arranged marriage compared with that of Kyrgyz. In addition, Table 3 shows that the risks of choice marriage are higher than those of arranged marriage among both ethnic groups; however, the corresponding difference is noticeably smaller among Uzbeks, a more traditional of the two groups.
The results of the joint modeling for ethnic-specific period changes in marriage risks are shown graphically in Fig. 1: the dynamics of both nonarranged and, especially, arranged marriage vary between Uzbeks and Kyrgyz. Particularly intriguing is the spike in risks of arranged marriage among Uzbeks in the early years of the post-Soviet era, which were characterized by a generalized socioeconomic crisis and a radical political transition. This result resonates with the findings on ethnic-specific demographic responses to economic hardships around the time of the collapse of the Soviet Union. Agadjanian and Makarova (2003:460), in their study of Uzbekistan, explained an increase in marriage risks during hard economic times by inflation of dowry, a marital payment practice that is particularly typical for arranged marriages among Uzbeks. They argued that the “. . . transformation of dowry requirements may have put pressure on families to marry their daughters earlier—before any further escalation of these requirements could take place.”
In contrast to Uzbeks, Kyrgyz experienced a decline in the risk of arranged marriage during the crisis years. Unlike the dowry-based Uzbek marriage, traditional marriage among Kyrgyz is primarily centered on bride price. To our knowledge, there are no sociological or anthropological studies that trace the evolution of bride price payments among Kyrgyz in the post-Soviet period. Although bride price inflation probably was also happening in that period, it was likely not as high as that of dowry given the lower cultural pressures to marry on men than on women. Importantly, the lesser traditionalism of Kyrgyz would also suggest that they should be more likely than Uzbeks to forego or renegotiate downward the marital payments. It is also interesting that the first two post-Soviet 5-year periods did not see any noticeable change in risks of nonarranged marriage among Kyrgyz: relative to the last Soviet decade, these risks declined insignificantly. In contrast, risks of nonarranged marriage increased considerably among Uzbeks in the early years after Kyrgyzstan’s independence. The last two periods under observation registered clear declines in all four ethnicity-marriage type categories. The decline was particularly dramatic for arranged marriages among Uzbeks, reversing the rise of the previous period. In 2006–2011, the risks of arranged marriage seemed lower among Uzbeks than among Kyrgyz; for Uzbeks, however, the corresponding result is based on a very small number of occurrences. In comparison, choice marriage risks among the two groups displayed very similar trends, declining at comparable rates between 2001–2005 and 2006–2011. It is noteworthy that for each ethnicity-marriage type category, declines were particularly pronounced between the period of late post-Soviet crisis and the period of initial economic recovery.
Multivariate Results: Marriages With and Without Bride Kidnapping
We now turn to the analysis of three competing transitions among ethnic Kyrgyz: to marriage without bride kidnapping, marriage involving mock bride kidnapping, and marriage involving forced bride kidnapping. Table 4 shows the results of separate models for each transition. In general, these results are consistent with those presented in Table 2: with few exceptions, they also detect similar associations between the covariates and marriage risks in the different marriage types. Most importantly, the estimates for period effects again indicate little change in marriage propensities between the late Soviet years and the start of the post-Soviet era, but a considerable decline in the two most recent periods. Decline in kidnapping-based marriages appear to start even earlier, although corresponding differences are not statistically significant. It is also noteworthy that the recent decline was particularly strong in the case of forced bride kidnapping.
Among other noteworthy results, women have significantly higher marriage risks compared with men. The gender difference is particularly pronounced in the model of marriage resulting from forced bride kidnapping, which is an indication of a considerable age gap between partners in marriages of this type. The effect of employment is not significant; it also appears to be negative in the model of forced bride kidnapping and positive in two other models. The effect of educational attainment is positive in the case of marriage without bride kidnapping and marriage involving forced kidnapping; the highest risk of mock bride kidnapping is found among respondents with intermediary educational level. Educational enrollment is negatively associated with marriage, irrespective of marriage type.
Next, we jointly model three competing transitions: to marriage without bride kidnapping, and to marriages resulting from mock or from forced bride kidnapping. Figure 2 illustrates graphically the period change in marriage risks. The results suggest a decline of marriage without bride kidnapping in the early post-Soviet era, followed by some recovery and then by a sudden drop. There were no changes in the risks of kidnapping-based marriages between the Soviet and first post-Soviet periods, but afterward the risks of these marriage types declined continuously. During the last decade under observation, there was a decline in all marriage types. In absolute terms, the overall decline is largest in the risks of marriages without bride kidnapping, but in relative terms, it is largest for marriages involving forced bride kidnapping. Thus, the risks of forced kidnapping were 77 % lower during 2006–2011 than during 1980–1990. For the two other marriage types, the declines between these periods constituted about 50 %. These results therefore show no increase in forced kidnappings in response to the post-Soviet socioeconomic crisis and the “lawlessness” of the transitional period as it sometimes is suggested in the literature. Also contrary to our expectation, the trends of mock kidnapping generally parallel those of forced kidnapping.
Summary and Conclusion
This study contributes to the limited research on nuptiality in Central Asia, a region of a unique historical mix of traditional and modern family-formation practices, with the special focus on Kyrgyzstan. We tested the relevance of the modernization and retraditionalization paradigms to contemporary marriage trends in this country by tracking the dynamics of traditional—arranged or forced—and choice marriages among its indigenous population. The analysis revealed that propensities of traditional and choice marriages changed only modestly between the late Soviet and early post-Soviet periods. The observed trends were not what one could have expected given the epochal socioeconomic and political shifts triggered by the demise of the USSR and by the massively increased openness of the region to both Western and Islamic cultural influences. However, the apparent nonresponsiveness of marriage dynamics to the generalized societal change in the first post-Soviet decade might be a result of counterbalancing effects of cultural modernization and retraditionalization. For the two most recent periods under observation, our analyses document a substantial decline of all types of marriages, arranged and nonarranged. Importantly, then, in this region, the economic rebound of the 2000s did not cause any recovery in marriage probabilities, as would have been expected from the pro-cyclical perspective on marriage response to economic swings. Although in conflict with much of the evidence from more conventional settings, these trends reflect the region’s specific historical circumstances and fit well with our modernization argument. Again, however, disentangling unique contributions of cultural, economic, and legal dynamics to these trends is impossible with the available data.
The comparative analysis of competing transitions by ethnicity, which related the risks of traditional marriage to the risks of choice marriage among Kyrgyz (the majority group with a relatively weak influence of Islam-rooted traditions) and Uzbeks (a minority with a stronger Islamic identity), revealed instructive differences. Most interestingly, the analysis detected a jump in the risks of arranged marriage and an increase in the risks of nonarranged marriage (even if less impressive than for arranged marriage) among Uzbeks during the first 5-year period after the dissolution of the USSR, while the corresponding risks among Kyrgyz slightly declined. The last post-Soviet decade under observation, however, showed a clear decline of marriage risks among both groups and in both types of marriages, with the decline being particularly steep in the risks of arranged marriages among the Uzbek minority. Over the entire observation span, arranged marriages declined faster than choice marriages, offering additional support to the modernization argument. Finally, the analysis demonstrated that the trends in forced and mock kidnappings among Kyrgyz paralleled each other, indicating that mock kidnapping is not a transitional form of marital union formation in that context.
Overall, the marriage trends in Kyrgyzstan detected in our study largely mirror those in other post-Communist settings. Studies in several such settings suggest that marriage rates somewhat increased there around the time of the collapse of the communist system and breakup of the USSR. The subsequent years, however, were characterized by considerable marriage declines. The declines did not stop after the most severe economic and sociopolitical shocks of the transitional period were over and continued, albeit at a slower pace, during the years of sociopolitical stabilization and economic recovery (e.g., Avdeev and Monnier 2000; Dommaraju and Agadjanian 2008; Philipov and Dorbritz 2003). Moreover, research has offered evidence that the massive societal cataclysms around the fall of communism notwithstanding, family formation in Eastern Europe and post-Soviet Eurasia generally conformed to a model charted by the second demographic transition theory, albeit with some context-specific nuances (e.g., Gerber and Berman 2010; Zakharov 2008). Although situated on the Muslim periphery of the former Soviet empire, Kyrgyzstan, then, appears to have followed a nuptiality pathway similar to that of the empire’s European successors. However, the noted differences between Kyrgyz and Uzbeks also illustrate the diversity in trajectories of marriage changes. Uzbeks, whose nuptial behavior may have been influenced by their deeper-rooted Islamic traditions as well as their minority status, displayed a seemingly anomalous increase in the probability of arranged marriage early in the post-Soviet era, while the titular ethnic group showed a moderate decline. Whether the explanation for an increase in marriage rates around the time of the collapse of the Soviet Union proposed by Agadjanian and Makarova (2003) for neighboring Uzbekistan is applicable to early post-Soviet trends in arranged marriage among Uzbeks in Kyrgyzstan remains open to debate. Yet, it is noteworthy that this early post-Soviet surge in arranged marriages reversed itself forcefully in the later post-Soviet periods, bringing the risks of arranged marriage among the two ethnic groups to comparably low levels despite the widely assumed greater traditionalism of Uzbeks. In fact, although we do not have data on earlier trends in arranged marriage in Kyrgyzstan, and the sample size and its age composition preclude an expansion and a more detailed periodization of the observation span, it would probably be safe to surmise that the levels reached in the first decade of the twenty-first century were the lowest in Kyrgyzstan’s history.
It should also be noted that the observed decrease in arranged marriage in Kyrgyzstan resembles the declines in arranged marriage documented in culturally diverse settings that had never been subjected to societal experiments comparable to the Soviet modernizationist project. In a number of East and Southeast Asian settings—where until recently, arranged marriage was an important part of the nuptiality system—this type of marriage all but disappeared (Jones 2010). In the settings where parent-arranged marriages still persist, their share has been shrinking rapidly. However, even though in such settings the share of choice marriages has been rising, the decline of arranged marriages has also been often accompanied by marriage postponement and increasing singlehood (Jones 1994, 2010; Jones and Gubhaju 2009; Retherford and Ogawa 2006; Retherford et al. 2004; Thornton et al. 1994).
The changes in marriage that we witness in Kyrgyzstan are part of the historical transformation of nuptiality behavior that is typically characterized by declining societal importance of marriage; the fading of its traditional types and functions; and as a result, marriage postponement and the spread of informal partnerships (Coontz 2004). This process is, of course, not linear nor straightforward but varies greatly across contexts and stages. Moreover, postponement of marriage and the proliferation of alternative forms of partnerships does not necessarily foretell the demise of the institution of marriage as recent trends in some Western settings attest (e.g., Ohlsson-Wijk 2011). An analysis of the consequences of this complex transformation in Central Asia is beyond the scope of our study. What our study does illustrate is how this process is locally manifested. Although anthropological and sociological accounts of unique and often dramatic expressions of traditional marriage practices provide useful insights into the real-life complexity of contemporary social fabric, they should not obfuscate the overall modernization of nuptiality in the region.
We thank Gunnar Andersson and Jan Hoem for their valuable comments on this article. The funding for the survey came from the National Council for East European and Eurasian Research (NCEEER), USA. The first author of this article is also grateful for financial support from Max Planck Institute for Demographic Research for the data analysis.
In our definition, “choice marriages” are not limited to marriages that result from decisions made solely and even primarily by the groom and the bride. In the context of Kyrgyzstan, as in many similar contexts, inputs from parents, other relatives, and even nonrelatives typically play important roles in most, if not all, marriages. Therefore, “choice marriages” are distinguished from “arranged marriages” by the fact of both the groom’s and the bride’s involvement in the marriage decision.
The oldest respondents in our data turned 16 in 1978, but there were no marriage occurrences before 1980. Our results might not be representative of the early 1980s because our sample has very few respondents who were in the peak marriage years at that time. Although this is a limitation, we believe that it does not affect the main results as variations in marriage risks during the last Soviet decade were probably minor.
We experimented with more categories for place of residence by additionally estimating risks for residence in Osh, Kyrgyzstan’s second-largest city, and subdividing urban and rural settlements into those located in the northern and southern parts of the country. The use of more nuanced classifications revealed no informative associations, and therefore we do not report corresponding results here (available upon request).
The effects of socioeconomic characteristics could possibly vary over time, reflecting, for instance, changes in gender-specific labor demands during the economic crisis and subsequent economic recovery. However, we leave this matter for future investigation because our data do not allow for a further splitting of the already small sample that would be required for an analysis of temporal trends across gender.