Bride Kidnapping and Gendered Labor Migration: Evidence from Kyrgyzstan

Abstract

Because the decision to migrate is a product of gendered negotiations within households, households formed through forced marriage may have different migration strategies than households formed through voluntary marriage. In Kyrgyzstan, we anticipate two possible effects of the traditional practice of bride kidnapping on migration. Households headed by a kidnap couple may be more cohesive and patriarchal, facilitating men’s labor migration and remittance-sending. Alternately, women may use migration to escape such households. We test these two hypotheses using a sample of 1,171 households in rural Kyrgyzstan. Kidnap households are more likely to include women migrants, compared to other households. Kidnap households are also more likely to be receiving remittances, even when controlling for migrant household members. However, traditional beliefs about kidnapping are negatively associated with men’s and women’s migration. While higher levels of remittance receipt among kidnap households resembles the unified, patriarchal households envisioned in the New Economics of Labor Migration, it also appears that women use labor migration as a means to escape patriarchal constraints. We demonstrate that forced marriage in Kyrgyzstan plays a larger social role than is often believed, and highlight a new pathway through which gendered power dynamics can shape household migration strategies.

Introduction

The decision to migrate is a product of complex, gendered negotiation processes within households. Research across a variety of cultural contexts shows that power dynamics within households shape the ways in which migration decisions are made, and who the household ultimately sends (or does not send) as a labor migrant (Cohen, Rodriguez, and Fox 2008; Hondagneu-Sotelo 1994; Nobles and McKelvey 2015). Household power inequalities arise from differences in control of household income and assets, the gender and age composition of the household, and cultural beliefs about gender, marriage, and stage of life. While power relations between spouses and other household members clearly matter for migration, we know little about how forced marriage potentially influences this dynamic.

Forced marriage is an alternative to the more common love and arranged marriages, in which the bride and groom have the opportunity to consent to their potential marriage partner, whether they select the potential partner themselves or the partner is selected by relatives or matchmakers (Becker, Mirkasimov, and Steiner 2017). We focus on a particularly dramatic form of forced marriage: bride kidnapping. Bride kidnapping describes a ritualized practice in which a woman is physically taken from her home or a public place, brought to the home of a prospective husband, and pressured into a marriage. Sometimes the pressure to marry comes from a threat of physical or sexual violence, but more often the social stigma of being kidnapped without marrying leads women to acquiesce to the kidnap marriage. Although bride kidnapping is not common on a global scale, it is practiced in a number of countries, including Kyrgyzstan, Kazakhstan, Armenia, Ethiopia, and South Africa (Bates, Conant, and Kudat 1974; Dito 2015; Edling 2012; Nkosi and Buthelezi 2013; Werner 2009). Notably, countries where bride kidnapping is practiced are often countries where labor migration (whether domestic or international) is also common, and it is likely that the processes of bride kidnapping and migration influence each other. Bride kidnapping may influence the gendered negotiation processes that drive household migration decisions in two ways. First, if a man forms his household by kidnapping his wife, it many indicate that he or other household members adhere to the traditional, patriarchal norms that have been associated with cohesive migration strategies in other contexts (Massey, Fischer, and Capoferro 2006; Sana and Massey 2005). Second, the possibility of being subject to a kidnap marriage may encourage young women to emigrate as a socially acceptable means of escaping patriarchal constraints.

We examine the relationship between bride kidnapping and migration using household survey data collected in 2017 in rural, montane Kyrgyzstan. Kyrgyzstan is notable for its extremely high prevalence of forced marriage. Although estimates vary, between 15 and 30% of all marriages are formed through a practice known locally as ala kachuu, which is usually translated as bride kidnapping (although sometimes as bride abduction or bride capture) (Nedoluzhko and Agadjanian 2015; Shaikhutdinov et al. 2015). Bride kidnapping also occurs elsewhere in Central Asia (particularly in Kazakhstan) but in Kyrgyzstan it is both more prevalent than elsewhere and widely embraced as a national tradition (Werner 2009). Although bride kidnapping was criminalized in Kyrgyzstan in the early 2000s, the practice persists, especially in rural areas (Kleinbach and Salimjanova 2007; Nedoluzhko and Agadjanian 2015; Shaikhutdinov et al. 2015). Both Kyrgyz and international researchers have analyzed the practice and meanings of ala kachuu, including the history of the practice, the extent to which it constrains women’s autonomy, and its relationship to discourses of nationalism. However, ways in which families formed through kidnapping may be different than families formed through voluntary marriage are only beginning to be explored (Becker et al. 2017; Steiner and Becker 2019).

Labor migration may be one aspect in which kidnap families are different from other types of families. In addition to its unique marriage practices, Kyrgyzstan experiences high levels of both men’s and women’s labor migration. After the collapse of the former Soviet Union, both international and domestic labor migration became common household survival strategies in Central Asia, with Kyrgyzstan ranking as the second most remittance-dependent country in the world in 2018, and neighboring Tajikistan ranking fourth (World Bank 2019). Kyrgyzstan stands out from its Central Asian neighbors due to its relatively high levels of women migration. While women comprise less than 25% of all international migrants from Tajikistan and Uzbekistan, they are as much as 40% of all migrants from Kyrgyzstan (Ahunov et al. 2015; Rocheva and Varshaver 2017). Understanding the relationship between Kyrgyzstan’s unique marriage patterns and its high levels of labor migration will further our understanding of household migration strategies in a context of forced marriage. In addition, we contribute to the debate over the role that bride kidnapping plays in 21^st century Kyrgyzstan and elsewhere.

Bride Kidnapping Practice and Family Dynamics in Kyrgyzstan

The script for a bride kidnapping in Kyrgyzstan is relatively standard (Borbieva 2012; Handrahan 2004; Werner 2009). The prospective groom and some friends either invite or force a woman into a waiting car. The woman is then taken to the prospective groom’s family home, where the women of the family try to convince her to put a white scarf on her head, a signal of her acceptance of the marriage. The kidnapped woman will resist for some time but—in most cases—will eventually put on the scarf. Rejecting the scarf and the marriage, although sometimes possible, carries significant social stigma for both the woman and her parents and can hurt the woman’s chances of a future marriage. If she accepts the scarf, the woman will then notify her parents, either directly or through the groom’s family, of her consent to the marriage. The bride’s and groom’s families will negotiate an agreement over the bride price, and the wedding can then take place.

This standardized script, however, conceals a great deal of diversity. Bride kidnappings sometimes occur with the woman’s prior consent. She may consent because she sees kidnapping as a romantic tradition, or she may consent in order to marry the man of her choice against the wishes of her parents. In such cases, the kidnapped woman may still feign resistance to the marriage, as a way of acting the part of a modest woman or dutiful daughter (Borbieva 2012). A 2015 public safety survey found that only 34 percent of bride kidnappings occur without the woman’s prior consent (Shaikhutdinov et al. 2015). However, another nationally representative study found that over half of bride kidnappings were non-consensual (Nedoluzhko and Agadjanian 2015). Studies by Kleinbach and colleagues, using non-random samples, found that 65 to 75 percent of kidnappings occur without the woman’s consent, and that some 20 percent of kidnappings involve rape (Amsler and Kleinbach 1999; Kleinbach 2003; Kleinbach, Ablezova, and Aitieva 2005). Kyrgyz scholars have argued that the line between consensual and non-consensual kidnappings is fuzzy; even when a kidnapping is arranged in advance, the bride may have agreed under duress, or may have little say in the timing of the marriage (Werner et al. 2018). The extent of involvement of the prospective groom’s family in a bride kidnapping also varies. Some grooms are pressured into kidnapping to secure a quick wedding, sometimes kidnapping occurs after a failed negotiation for an arranged marriage, and sometimes young men kidnap without the prior knowledge of their families (Amsler and Kleinbach 1999; Borbieva 2012; Kleinbach et al. 2005; Werner 2009).

Scholars have debated the cultural meanings of bride kidnapping. Werner (2009) describes it as a sign of resurging patriarchal norms in post-Soviet Central Asia. Kleinbach and colleagues (2005) argue that it constitutes a failure of the patriarchal norm of arranged marriage. However, Nedoluzhko and Agadjanian (2015) test the hypothesis that bride kidnapping, particularly consensual kidnapping, serves as a replacement for traditional arranged marriages and find no evidence for it. Summing up the complexity of the issue, Borbieva (2012: 141) describes bride kidnapping as “a practice that both supports and undermines existing systems of oppression.”

Kidnapping is not universally common across Kyrgyz society. It is substantially more common in rural than in urban areas, and it is concentrated among ethnic Kyrgyz and almost unheard of among other ethnic groups (Becker et al. 2017; Nedoluzhko and Agadjanian 2015). It also appears to be related to educational levels; men with university-level education rarely kidnap, with kidnaping most common among men with intermediate levels of education (Becker et al. 2017). Kidnapping is also associated with men of lower social status, whether that status refers to the wealth and influence of their family of origin, or to their own charm and social skill (Werner et al. 2018). Kidnapping selectively affects women as well, with ethnographic research showing that women perceived to be shy or docile are seen as ideal targets for kidnapping, and that women whose families openly discuss how to respond in the event of a kidnapping are somewhat protected from kidnapping (Werner et al. 2018).

The prevalence of bride kidnapping is changing over time. Despite criminalization of the practice in the early 2000s, survey data show bride kidnapping to be consistently common until around 2010, when it decreased somewhat (Becker et al. 2017; Nedoluzhko and Agadjanian 2015). The decline may be accelerating in recent years due to increasing local activism and stricter criminal penalties established in 2012 (Werner et al. 2018). Nevertheless, as recently as 2019, a popular television program portrayed bride kidnapping in a positive, romantic light (Furlong 2019).

Little is known about how kidnapping affects family dynamics, or how kidnap marriages are different from other marriages. In the limited research on forced marriage in other countries, it is negatively associated with psychological wellbeing (Ali et al. 2009; Samad 2010), but bride kidnapping in Kyrgyzstan is often perceived as largely symbolic. However, there is growing evidence against the belief that bride kidnapping in Kyrgyzstan is just a local ritual that is practically equivalent to an elopement. Becker and colleagues (2017) found that children born to mothers in kidnap marriages showed lower birth weights than children born to mothers in other types of marriages. Although they were not able to identify mechanisms through which kidnapping leads to lower birthweight, they hypothesized that lower levels of homogamy and higher stress levels among kidnapped wives are likely culprits. In related research, Steiner and Becker (2019) showed that kidnap marriages demonstrate less similarity between husbands’ and wives’ personality types. Neither of these studies was able to distinguish between coercive and staged kidnappings, so the authors conclude that their estimates of the effects of kidnapping may be conservative.

Migration and Family Dynamics

Research from a variety of migrant-origin countries demonstrates that characteristics of households shape multiple aspects of migration, including whether households are able to use labor migration as a risk-management or investment strategy, and whether men or women dominate migrant flows. Below we discuss these aspects within the Kyrgyz context of bride kidnapping and gendered labor migration.

Household Migration Strategies

Labor migration is often conceptualized in the literature as a household-level strategy to diversify income sources, earn money for investment, and minimize financial risks—the perspective outlined in the New Economics of Labor Migration (NELM) (Massey et al. 1998; Stark and Bloom 1985). NELM has been criticized for treating households as the decision-making unit and ignoring power dynamics, especially gendered power dynamics, within households (Hondagneu-Sotelo 2003; King 2007). Sana and Massey (2005: 525) argue that the NELM theory of migration is most relevant in societies such as Mexico where the typical family is “cohesive and, owing to patriarchal norms, likely to honor contractual arrangements that ensure flows of remittances from family members abroad.” Remittance sending is often tied to migrants’ gendered obligations as sons, daughters, husbands, wives, mothers, and fathers (Carling 2014; Park, Cruz-Saco, and Anuarbe 2017; Wong 2006), and household heads may deliberately choose family members with the strongest sense of family obligation to be given encouragement or permission to migrate (Curran et al. 2005). Cross-national research highlights that the migration and remittance patterns described by NELM are most typical in countries where stable, patriarchal families are the norm. In countries with less stable or more egalitarian families, there is less remittance sending, and migration is often more of a move of desperation than a strategic household plan (Massey et al. 2006; Oishi 2005; Sana and Massey 2005).

Remittance sending is an important aspect of migration from Kyrgyzstan, which receives high levels of remittances relative to its population and GDP. Migration from Kyrgyzstan has other similarities to the pattern described by NELM as well. The typical migrant from Kyrgyzstan is a an unmarried young adult from a rural area (Olimova and Olimov 2007; Sagynbekova 2017; Vinokurov 2013), but data from our field work suggest that married couples with young children comprise a notable share of migrants. Rural Kyrgyzstan is characterized by large, multigenerational and patrilocal households. Youngest sons are expected to remain permanently in the parental home (which they will eventually inherit), but even older sons often live with their parents for at least a few years after marriage. Women either move in with their husbands’ parents or form a separate, nuclear household after marriage; they almost never continue to reside with their own parents (Landmann, Seitz, and Steiner 2018). In other words, Kyrgyzstan has precisely the type of household that should facilitate the use of labor migration as a household-level economic strategy.

But Kyrgyzstan differs from the standard NELM narrative in one important respect. The cross-national research arguing that patriarchal families facilitate remittance flows also finds that the same family type facilitates male-dominated migrant flows (Massey et al. 2006; Oishi 2005; Sana and Massey 2005). Ethnographic research confirms that in Mexico, men frequently migrate without even consulting their wives or other women in their household, while women migrate primarily as tied migrants following male relatives (Broughton 2008; Cohen et al. 2008; Hondagneu-Sotelo 1994). Kyrgyzstan, however, has an unusually high level of women’s migration among Central Asian states, with unmarried young women nearly as likely to migrate as unmarried men (Rocheva and Varshaver 2017; Thieme 2008). This difference is often attributed to higher levels of women’s autonomy in Kyrgyzstan, compared to neighboring Uzbekistan or Tajikistan (Rocheva and Varshaver 2017).

While the prevalence of patriarchal norms and patriarchal families seems to predict national-level migration patterns, households within a country are not uniform with regard to their adherence to patriarchal norms. In the Republic of Georgia, women migrants are more likely to originate from households where adult household members express gender-egalitarian views. A history of men in the household performing domestic chores is associated with a higher likelihood of migration among women, and a lower likelihood among men (Hofmann 2014). Using data from a policy experiment in Mexico, Nobles and McKelvey (2015) found that when women have greater control over household resources, their husbands are less likely to migrate. They argue that Mexican women are less attracted to the idea of migration than men are, and that men’s migration is therefore most common from households in which women’s voices carry the least weight (Nobles and McKelvey 2015). Women’s relationship power, in short, can be directly predictive of both men’s and women’s migration. However, research on household head gender views and women’s labor force participation has found no relationship (Landmann et al. 2018), so the relevance of these theories to Kyrgyzstan is unclear.

Migration and household conflict

The decision of a household member to migrate may be the result of not just negotiation but outright conflict. Qualitative research clearly shows that people sometimes migrate against the expressed wishes of their parents or spouses (Asis, Huang, and Yeoh 2004; Hondagneu-Sotelo 1994; Tucker et al. 2013). Sometimes, apparent agreement on migration can mask very different reasoning, such as when the household head sees migration as an investment strategy, while the migrant sees it as an escape from dependence (Grasmuck and Pessar 1991). In other cases, migration can be a unilateral decision on the part of the migrant to escape an intolerable situation at home (Pedraza 1991; Ryan et al. 2009).

Sometimes household conflict is apparent before the migrant leaves, but other times it appears only later, when the migrant decides to remain in the destination country, or to cease remitting against the wishes of their family. The risk of a migrant forming a new sexual partnership in their destination country, and turning away from their original family, is well documented (Van Hook and Glick 2020). Although less well established, there is also evidence that women in particular often come to appreciate the increased autonomy they experience as labor migrants, whether from the independence that comes from earning a salary, the experience of living in a new place with a more egalitarian culture, or simply the physical distance from watchful family members (Dannecker 2005; Foner 2002; Guendelman and Perez-Itriago 1987; Torosyan, Gerber, and Goñalons-Pons 2016). This increased autonomy, in turn, is associated with a higher desire among women to remain permanently in their destination (Pessar 1999).

Hypotheses

The existing literature leads us to two hypotheses on the relationship between gender, bride kidnapping, and migration in Kyrgyzstan, which are detailed in Figure 1. The first, which we refer to as the patriarchal norms hypothesis, assumes that men who kidnap their wives have a strong attachment to patriarchal norms, and that this influence is felt throughout their household. Because strong patriarchal norms make a household more likely to use (male) labor migration and remittances as an investment strategy, we would expect that households in which the head married by kidnapping would be more likely to include migrants, and particularly more likely to include male migrants, compared to households without experience of kidnapping. In addition, we would expect those migrants to be more likely to send remittances, because they migrate in response to their household head’s wishes, rather than for their individual benefit. Finally, if the patriarchal norms hypothesis is accurate, we would expect that the positive relationship between bride kidnapping and migration would be at least partly explained by adding control variables that measure the attachment of household members to traditional social or gender norms.

Figure 1.

Conceptual Model and Hypotheses

The second hypothesis, which we refer to as the escape hypothesis, assumes that, instead of kidnap households being the most cohesive and able to use labor migration as an investment strategy, these are the households from which young people are most inclined to escape. The desire to escape should be strongest among young women. Unmarried women are best protected against kidnapping when their relatives have both the desire and the social standing to confront the kidnapper and his family (Werner et al. 2018). A household head who himself married through kidnapping would presumably be less inclined to confront a man who kidnapped his daughter; labor migration might therefore be a socially legitimate means of escape for a young woman who wanted to avoid the possibility of a kidnap marriage. If the escape hypothesis is accurate, we would expect to see that households in which the head married by kidnapping are more likely to include migrants, but that this relationship is strongest for female migrants. But since this migration is more of an escape than an investment strategy, we would not see the higher probability of remittances assumed by the patriarchal norms hypotheses. We would expect the positive relationship between kidnapping and migration posited by the escape hypothesis to be at least partially explained if we control for household members’ beliefs about the stigma of refusing a kidnap marriage.

Data

Data to test these hypotheses come from a household survey conducted in a rural, high-altitude region of Kyrgyzstan over a two-week period in the summer of 2017. A total of 1,233 households were surveyed in 13 villages of the Alay Rayon (County) of Osh Oblast (Province). We selected the villages based on a stratified random procedure with variation by (1) village category (rayon city, municipality village, and village), (2) population size, (3) elevation, and (4) remoteness. The village populations range from 224 to 11,691 people as of 2017. The village elevation ranges from 1,482 to 3,362 meters above sea level. Although Kyrgyzstan is a multi-ethnic country, this particular rayon is almost entirely ethnic Kyrgyz, and our sample is therefore monoethnic and the survey was done in the Kyrgyz language only. Surveys were conducted by a team of 10 local school teachers serving as interviewers. When the interviewers arrived in a village, they were split into pairs (due to safety concerns) and each pair took an entire street or a segment of it to conduct surveys. They went door by door and conducted a survey when an adult was at home and was willing to take the survey. The data collection was reviewed and approved by the Institutional Review Board (IRB) of Pennsylvania State University.

The surveys included a wide variety of questions about demographics, socioeconomic statuses, migration, remittances, land use and agriculture, and wellbeing. The instrument directly asked both marital status and whether the marriage in question was a kidnap marriage or not—an approach that has been successful in previous survey research in Kyrgyzstan (Becker et al. 2017; Nedoluzhko and Agadjanian 2015). In the survey, 26% of reported marriages were kidnap marriages, a prevalence rate similar to those reported by Becker et al. (2017) and Nedoluzhko and Agadjanian (2015), but higher than reported by Shaikhutdinov et al. (2015).

The household member chosen as the respondent was an adult at home during the interview. When multiple adults were at home, the survey proceeded with the typical order of status in the Kyrgyz family culture—grandfather, grandmother, father, and mother. The respondent provided reports on all children in the household. The respondent also provided proxy information on all household members who were currently living as migrants either elsewhere in Kyrgyzstan or abroad. The definition of a “household member” was deliberately open-ended in the survey, in order to capture as many migrants as possible. Respondents were asked to list people who live and work elsewhere, but “remain connected and part of your household” (in Kyrgyz, “Сиз менен байланышта турган жана үй-чарбаңыздын мүчөсү болгон адамдар жөнүндө болмокчу”). Proxy reports tend to underestimate migration. When a whole household migrates, leaving no member behind to be sampled, those migrants are missed, and they likely have different characteristics than migrants who leave family members behind. Nevertheless, proxy reports are commonly used to measure both prevalence and characteristics of out-migration, especially in the former Soviet Union (Atamanov and Van den Berg 2012; Makaryan 2015; Thomas-Hope et al. 2009; Torosyan et al. 2016).

Methods

Our analytic sample includes 1,175 households comprised of two or more individual members (including migrants). Fifty-eight single-person households were excluded from our original sample of 1,233 because by definition, a one-person household could not include a migrant. Although single-person households by our definitions could report remittance receipt, for the sake of comparability we excluded them from all analyses. Item non-response in the survey was low. Only 72 households (6%) had missing values on any variables of interest (most commonly household income).

We used both migration and remittance receipt as dependent variables. Remittance receipt was operationalized as a simple dichotomous variable, taken from a list of the primary sources of household income provided by the respondent. Household-level migration was operationalized as a three-category variable: households that reported no migrants; households that reported at least one male migrant but no female migrants; and households that reported at least one female migrant or both male and female migrants. Because households can include more than one migrant (of the 173 households that include migrants, nearly 40% include multiple migrants), this is not the only possible categorization of household migration strategy. Households that include both male and female migrants, including those from which a married couple migrates together, were classified as female migrant households because our sensitivity testing showed that they were most similar to households with only female migrants. Sensitivity testing also showed little difference between migrants who traveled solo (the majority of migrants in our sample) versus those who traveled as couples, so we did not distinguish between solo and couple migration.

We measured the effects of kidnapping at the household level by coding households in which the household head (oldest male household member, or oldest woman in the absence of an adult man) married by kidnapping. This includes both current and former marriages. In other words, a household in which the head is widowed or divorced was coded as a kidnap household if the head described their former marriage as a kidnap marriage. Because women almost never continue to live with their parents after marriage in Kyrgyzstan, a kidnap household can be either a nuclear family or a multigenerational household that includes the kidnapping husband (and his wife if they remain married) as well as their adult children, children-in-law, and grandchildren.

We included the age, age-squared, and age-cubed of the household head because the prevalence of both kidnap marriage and migration in Kyrgyzstan have varied substantially over time. We controlled for household size since larger households are more likely to include migrants. Because education level is related to both bride kidnapping and migration, we included a variable for the educational level of the household head, split into three categories: high school education or less (very few people in the sample had not finished high school); some post-secondary education; and college degree or above. We also tested a variable for highest education level of any household member, but this was highly correlated with the education level of the household head and produced nearly identical results.

Household socioeconomic status influences both propensity for migration and whether remittances are a part of the household’s economic strategy. We included dummy variables for whether the household receives any income from a household member’s salary, from agricultural activities (including livestock husbandry), and from government transfers (most commonly old-age pensions). In addition to categorizing sources of income, we included a variable for the household’s reported monthly income (in 1000s of Kyrgyz soms, approximately 12 USD), divided by the total number of household members. We also included a dummy variable for whether or not members of the household own any land.¹

Finally, we included a number of control variables for the beliefs and values reported by the primary respondent (usually the household head). The survey collected Likert scale responses (with higher values indicating stronger agreement) to a number of statements about personal and community norms, including: bride kidnapping is a common practice in my community; divorce has a negative stigma in my community; big weddings are very important; refusing a kidnap marriage has a negative stigma in my community; domestic violence is common in my community. We created a scale from these five variables, with an alpha value of 0.75. We also created dichotomous variables indicating that the respondent agreed or strongly agreed with each of the statements.

We conducted multinomial logit models predicting the odds of having men-only migrant family members and migrant family members including women, relative to having no migrants in the household. We also conducted logistic regression models predicting the odds of receiving remittances. As a robustness check, we ran propensity score models matching kidnap and no kidnap households on the set of control variables. Results were the same, so only the logit models are presented here. Missing values for per capita household income and age of household head were imputed using data augmentation with an assumed multivariate normal distribution (Little and Rubin 2019). All models were run using the mi impute mvn command in Stata 14 (StataCorp 2015), and standard errors were adjusted for clustering at the village level.

Descriptive Results

Survey data show limited differences between households based on whether or not the head married by kidnapping (see Table 1), with the most striking differences being in migration behavior and reported beliefs and norms. Although having a current migrant household member and reporting significant remittance income are both fairly uncommon in this sample (fewer than 20% of all households), kidnap households are notably more likely to include both male and female migrants, compared to other households. The difference in remittance income is even greater, with kidnap households more than twice as likely as other households to report remittances (10.47% versus 4.61%).

Table 1.

Characteristics of households by household head’s kidnapping status

	Head not in kidnap marriage	Head in kidnap marriage
Migration characteristics
% no migrants in household	87.28	80.97
% only male migrants in household	5.24	7.51
% at least one female migrant in household	7.48	11.53
% receiving remittances	4.61	10.47
Mean household size	4.38	4.82
Mean age of household head	46.31	43.05
Household head education level
% high school or less	64.21	69.44
% some post-secondary	20.45	16.35
% college degree	15.34	14.21
% households having any salary income	33.17	32.44
% households having any agricultural income	44.01	42.63
% households receiving any government assistance	40.77	43.43
Mean per capita household income	23.34	20.03
% owning land	58.95	59.95
Mean household norms index	4.03	3.85
% agreeing that kidnapping is common	19.70	28.69
% agreeing that rejecting kidnapping carries stigma	25.56	34.05
% agreeing that divorce carries stigma	28.68	34.05
% agreeing that big weddings are important	26.18	37.53
% agreeing that domestic violence is common	15.46	15.82
N	802	373

Household heads who married by kidnapping are slightly younger than household heads who are not. This difference is likely related to both the especially high prevalence of kidnapping in the early 2000s, and the fact that couples in kidnap marriages are more likely to form separate nuclear households rather than reside with the groom’s parents. Kidnap households are also larger than other households, with average household sizes of 4.82 and 4.38, respectively. Although not shown in the table, the larger size of kidnap households is mostly explained by larger numbers of minor children in the household, providing some support for our contention that kidnap households are more traditional. However, none of these demographic differences between households are large. Kidnap households have slightly lower socioeconomic status than other households, in terms of both education level and household income. Kidnap households are more likely to have a head who has no education beyond high school, and the also have per capita incomes of about 2000 KGS (about $25) less per year than other households. They are also more likely to be receiving government support.

Differences in beliefs and norms are somewhat larger, and generally in the expected direction. When the household head married by kidnapping, the primary respondent is more likely to report that kidnapping is common in the community, and that refusing a kidnap marriage is socially stigmatized. They are also somewhat more likely to report that divorce is stigmatized, and that big weddings are important. There is no difference in beliefs about domestic violence, and the difference in the overall household norms index between household types is small. The relevant survey questions were phrased to measure the respondent’s perception of their community as a whole, rather than their own personal beliefs. However, there is almost no correlation between village of residence and answers to any of the community norms questions, indicating that the response is more of a reflection of the respondent’s perspective than a description of the community.

Multivariate Models

The descriptive statistics provide some support for both the patriarchal norms and escape hypotheses. Men’s migration and remittances are both more common in kidnap households, as the patriarchal norms hypothesis predicts, but women’s migration is equally associated with kidnapping, which is more in line with the escape hypothesis. We turn to multivariate models to investigate the role that socioeconomic and demographic differences between households have in explaining the kidnapping-migration association, as well as to examine the possibility that controlling for household norms attenuates the association.

Tables 2 and 3 show the results of multinomial logit models predicting log odds of a household including either male migrants only or any female migrants (relative to no migrants). For ease of presentation, log odds of having male migrants only is presented in Table 3, while log odds of having any female migrants are presented in Table 4. In both tables, Model 1 includes demographic and socioeconomic control variables, Model 2 adds the household norms index, and Model 3 includes controls for each norms question separately.

Table 2.

Estimated coefficients and standard errors (in parentheses) of multinomial logit models predicting presence of male migrants (relative to no migration)

	Model 1	Model 2	Model 3
Head in kidnap marriage	.261 (.269)	.251 (.271)	.300 (.277)
Household size	.502*** (.068)	.500*** (.068)	.492*** (.070)
Age of household head	−.624** (.187)	−.619** (.187)	−.600** (.193)
(Age of household head)2	.013** (.004)	.012** (.004)	.012** (.004)
(Age of household head)3	−.000** (.000)	−.000** (.000)	−.000** (.000)
Head’s education level (ref=high school or less)
Some post-secondary	−.090 (.370)	−.102 (.371)	−.224 (.374)
Completed tertiary	.080 (.367)	.068 (.370)	.118 (.378)
Any salary income	.088 (.283)	.090 (.282)	.042 (.289)
Any agricultural income	−.107 (.273)	−.133 (.286)	−.305 (.290)
Any government support	−.597* (.301)	−.587† (.301)	−.771* (.315)
Per capita household income	.001 (.003)	.001 (.002)	.000 (.003)
Owns land	.671* (.297)	.669* (.297)	.627* (.303)
Household norms index		−.058 (.161)
Kidnapping is common			−1.206** (.448)
Refusing kidnapping is stigmatized			−.458 (.410)
Divorce is stigmatized			−.258 (.395)
Big weddings are important			.715* (.298)
Domestic violence is common			.051 (.424)
Constant	4.044	4.183	3.951
F statistic	7.70	7.11	5.82
N	1,171	1,171	1,171

^†p<.10

^*p<.05

^**p<.01

^***p<.001

Table 3.

Estimated coefficients and standard errors (in parentheses) of multinomial logit models predicting presence of female migrants (relative to no migration)

	Model 1	Model 2	Model 3
Head in kidnap marriage	.438† (.249)	.397 (.251)	.422† (.257)
Household size	.578*** (.065)	.577*** (.065)	.565*** (.067)
Age of household head	−.945*** (.261)	−.928*** (.262)	−.934** (.269)
(Age of household head)2	.023*** (.006)	.023*** (.006)	.023*** (.007)
(Age of household head)3	−.000*** (.000)	−.000*** (.000)	−.000*** (.000)
Head’s education level (ref=high school or less)
Some post-secondary	.547† (.327)	.500 (.327)	.266 (.378)
Completed tertiary	1.376*** (.338)	1.28*** (.344)	1.251*** (.357)
Any salary income	−.621* (.302)	−.624* (.302)	−.698* (.311)
Any agricultural income	−.464† (.261)	−.611* (.275)	−.798** (.282)
Any government support	−.355 (.277)	.390 (.279)	.260 (.292)
Per capita household income	−.014* (.007)	−.014* (.007)	−.015* (.007)
Owns land	1.162*** (.300)	1.173*** (.301)	1.068** (.308)
Household norms index		−.259† (.154)
Kidnapping is common			−.794* (.373)
Refusing kidnapping is stigmatized			−.654† (.383)
Divorce is stigmatized			.465 (.349)
Big weddings are important			1.097*** (.288)
Domestic violence is common			−.456 (.447)
Constant	5.179	4.183	4.972
F statistic	7.70	7.11	5.82
N	1,171	1,171	1,171

^†p<.10

^*p<.05

^**p<.01

^***p<.001

Table 4.

Estimated coefficients and standard errors (in parentheses) of logistic regression models predicting remittance income

	Model 1	Model 2	Model 3
Head in kidnap marriage	.800** (.282)	.725* (.287)	.761** (.288)
Household migrants (ref=no migrants)
Male migrants only	3.380*** (.374)	3.390*** (.375)	3.380*** (.386)
Any female migrants	3.001*** (.382)	2.968*** (.383)	2.900*** (.392)
Household size	.078 (.065)	.076 (.066)	.058 (.066)
Age of household head	−.625* (.252)	−.634* (.257)	−.654* (.262)
(Age of household head)2	.014* (.006)	.014* (.006)	.015* (.006)
(Age of household head)3	−.000* (.000)	−.000* (.000)	−.000* (.000)
Education level (ref=high school or less)
Some post-secondary	−.097 (.378)	−.151 (.381)	−.200 (.385)
Completed tertiary	.498 (.380)	.391 (.388)	.388 (.392)
Any salary income	−.403 (.330)	−.423 (.331)	−.467 (.341)
Any agricultural income	.444 (.291)	.314 (.303)	.280 (.303)
Any government support	−.331 (.317)	−.301 (.317)	−.338 (.325)
Owns land	−.625* (.317)	−.627* (.219)	−.700* (.322)
Household norms index		−.293 (.194)
Kidnapping is common			.115 (.408)
Refusing kidnapping is stigmatized			.173 (.429)
Divorce is stigmatized			−.071 (.411)
Big weddings are important			.682* (.325)
Domestic violence is common			−.613 (.529)
Constant	4.158	5.383	4.367
F statistic	10.55	9.73	7.52
N	1,171	1,171	1,171

^†p<.10

^*p<.05

^**p<.01

^***p<.001

A first key finding of the tables is that, while the bivariate relationship between kidnapping and men’s and women’s migration is similar, the relationship between kidnapping and women’s migration is more robust to the inclusion of sociodemographic control variables. The household head being in a kidnap marriage is positively associated with both men’s and women’s migration across all models, but the coefficient for men’s migration is smaller and never achieves statistical significance in any model. The larger size of kidnap households appears to explain much of their greater presence of male migrants. This is somewhat true for female migrants as well, but the positive effect of being a kidnap household on women’s migration is larger across all models, and significant in both Models 1 and 3 (and almost so in Model 2). Kidnapping therefore seems to have more influence on women’s migration than on men’s.

There are also notable differences between the variables that predict men’s and women’s migration. Household size is positively associated with both men’s and women’s migration, and both have a similar relationship with the household head’s age, in which migration is most common when the household head is either very young (early 20s) or middle aged (40s), but decreases at other ages. Household education and income, on the other hand, affect men’s and women’s migration differently. Men’s migration is most common where the household owns land and does not receive government support, with the education level of the household head mattering little. A more highly educated household head has a strong positive relationship with women’s migration, while salary income, agricultural income, and a higher per capita household income ² all lower the chances of women’s migration.

The second major finding from Tables 2 and 3 is that household beliefs and norms are associated with household migration strategies, but that they do not explain the association between kidnapping and migration. Model 2 shows that higher scores on the household norms index (indicating more conservative or traditional values) has no influence on men’s migration, but a negative influence on women’s migration. Looking at specific beliefs separately in Model 3 shows that the relationship between beliefs and migration is not as simple as a conservative versus egalitarian scale. Belief in the social importance of big weddings is positively associated with having both male and female migrants. This association is not surprising; wanting to host a large (and expensive) wedding might be a motivation to engage in migration in the first place, or the extra income provided by migrant remittances could make a household more interested in the idea of a large wedding.

Belief that bride kidnapping is common in the community has a fairly strong negative relationship with both men’s and women’s migration, and women’s migration is also negatively related to the belief that women cannot walk away from a kidnapping without facing social stigma. Beliefs about divorce and domestic violence are not significantly associated with either men’s or women’s migration. The inclusion of norms measures does not substantially attenuate the relationship between bride kidnapping and migration in any model.

The findings presented in Tables 2 and 3 support the escape hypothesis, although not unequivocally so. The positive association between kidnapping and women’s (but not men’s) migration is clearly in line with the escape hypothesis. However, beliefs about the prevalence of kidnapping not only fail to explain the higher levels of women’s migration, they actually contradict it. Belief that kidnapping is both prevalent and risky is associated with lower levels of women’s migration, and men’s migration shows a similar relationship. Living in a kidnap household is therefore not just a proxy for stronger traditional gender norms, or at least is reflecting different norms than the limited gender norms available in this survey.

Table 4 displays the results of logistic regression models predicting remittance receipt, including the same independent variables that were included in the models above. However, the models in Table 4 also include controls for the migration status of the household. Although having migrant household members and receiving remittances are strongly correlated, not all migrants send remittances, and some households without migrants nonetheless receive remittances (for example, from an extended family member whom the respondent did not consider as a household member). Bivariate analyses (see Table 1) showed that more than twice as many kidnap households received remittances as did non-kidnap households. The multivariate analyses in Table 4 show that higher levels of remittance receipt are not simply because kidnap households include more migrants. Not surprisingly, having a migrant in the household is by far the strongest predictor of receiving remittances, and this association is very similar whether the migrants are men or women. However, a fairly strong association between living in a kidnap household and receiving remittances remains across all three models. Furthermore, the coefficients for kidnapping have among the lowest p-values across all models. Similar to the previously discussed models in Tables 2 and 3, measures of beliefs and norms do almost nothing to explain the association between kidnapping and remittance receipt. The only norms measure that has any significant association with remittance receipt is the belief that big weddings are important, which is unsurprisingly associated with a higher likelihood of remittances.

The age of the household head is a significant predictor of remittances, showing a similar cubic relationship to that demonstrated in the previous tables, in which very young and middle-aged household heads are the most likely to be receiving remittances. There are few other significant predictors of remittances in the model, although the coefficients, even when not significant, are generally in the directions predicted by NELM. If the household receives income from a member’s salaried job or from government benefits such as an old-age pension, they are less likely to receive remittances. This finding is consistent with the NELM contention that remittances serve as an insurance mechanism against unpredictable income streams (Stark and Bloom 1985). Because pensions and salaries are the most reliable sources of regular income in rural Kyrgyzstan, households with this income available have less need for remittances. Having income from agricultural activities, on the other hand, is positively associated with receiving remittances, while owning land is negatively associated with remittances. This again is consistent with NELM predictions—households that depend on agriculture for their livelihoods, but do not own land, have a strong incentive to receive remittances in order to purchase land or at least ensure that they can pay rent on the land that they use.

Overall, the results of Table 4 provide the strongest support for the patriarchal norms hypothesis. The results, although not conclusive, are consistent with the idea that remittances in Kyrgyzstan are used as part of a household-level economic strategy, and that kidnap households are more likely to use remittances in this way, regardless of their land ownership or other income sources. There is no evidence that migration from kidnap households functions as a method of escape rather than as a part of the household’s economic strategy. However, just as in the models presented in Tables 2 and 3, if kidnapping serves as an indicator of a particularly patriarchal household, this is not reflected in the gender beliefs measured by the survey.

Conclusion and Discussion

This study was initially motivated by the question: Do households formed through a kidnap marriage engage in different migration strategies than other households do? Survey data from the Alay region in rural southern Kyrgyzstan provide a clear affirmative answer. Households headed by a man or woman who married by kidnapping are more likely to include current female migrants, and are more likely to receive remittances. Kidnapping was in fact one of the strongest associations with reporting remittances, with a more significant association than most of the available socioeconomic variables.

Answers to the follow-up question of why kidnap households might have different migration strategies are more elusive. We found partial support for both the hypothesis that women use migration to escape from either an existing kidnap marriage or the possibility of being forced into such a marriage in the future, and for the hypothesis that patriarchal norms facilitate both labor migration and remittances. In multivariate models, the positive association between kidnap marriage and migration was only significant for women’s migration, as predicted by the escape hypothesis. The remainder of the escape hypothesis, which stated that the association between kidnapping and women’s migration would be attenuated by including measures of household norms, and that migrants from kidnap households would be less likely to send remittances, received no support. The patriarchal norms hypothesis was also only partially supported. While there was no support for our prediction that the association between kidnapping and migration would be driven primarily by male migration, there was strong support for the prediction of higher rates of remittance receipt among kidnap households.

This research has a number of important limitations that may be driving these conflicting findings. First, the survey included limited information about marriage and migration. We know whether or not the household head’s current or most recent marriage was a kidnap marriage, but we cannot distinguish between levels of coercion in kidnapping, and we do not know the timing of the marriage. In addition, if the household head had a previous kidnap marriage and then remarried without kidnapping, we have no record of the original, kidnap marriage. Our data may therefore underreport the prevalence of kidnap marriage. However, given that our estimates of kidnapping prevalence are similar to those of other studies, we do not think that this underreporting is substantial. We also lack information on previous migration. Our definition of households with no migrants only means no current migrants; members of the household may well have migrated in the past, although we have no reason to suspect that this undercount would affect kidnap households any differently than other households. A second, and related, limitation is that the cross-sectional data do not allow us to make causal inferences. Although similar research often employs an instrumental variable approach, we were unable to identify a suitable instrument to model the relationship between kidnapping and migration.³ The lack of data on marriage or migration timing was a particular stumbling block to identifying an instrument. Selection into a kidnap marriage (whether as the kidnapping groom or the kidnapped bride) is by no means random, and we cannot rule out the possibility of unobserved heterogeneity.

A final limitation is the household norms measures that were available. Although we believe that variation in the beliefs and norms measures reflects variation in the beliefs and opinions of the household respondent, it is not certain that a “traditional” response to the beliefs questions actually represents a traditional worldview. For example, a respondent who holds more egalitarian beliefs, and perceives domestic violence or bride kidnapping as a serious problem, might be more likely to identify those issues as prevalent in the community. Someone who is not concerned over these issues might be less attuned to them and therefore perceive them as less prevalent. In our data, women who respond to the survey on behalf of their household are significantly more likely to describe kidnapping as common, and to say that refusing a kidnapping is stigmatized, compared to male respondents. Controlling for respondent gender did not improve our models, but does demonstrate that these norms measure are far from objective. These measures of household norms do show a clear relationship with migration, especially women’s migration, but more research will be needed to understand why this relationship exists.

These limitations notwithstanding, one possible explanation for our mixed findings is that the processes described in the patriarchal norms and escape hypotheses are both happening simultaneously. Because labor migration is widely practiced and culturally normative in rural Kyrgyzstan, people with vastly different motivations may pick the same strategy. Remittance-sending, for example, may be either a household strategy demanded by a dominant patriarch, or way for a young woman to appease or “pay off” her parents if she defies their wishes and looks for work abroad.

The findings of this study have two key implications. First, our findings provide further support for the idea that families formed through forced marriage are fundamentally different than families formed through other types of marriage. The limited research on forced marriage has shown negative effects on psychological wellbeing (Ali et al. 2009; Samad 2010), but bride kidnapping in Kyrgyzstan is often perceived as largely symbolic. When Human Rights Watch (2006) interviewed Kyrgyz government officials, they found that the officials viewed kidnappings as mostly consensual and unproblematic. These attitudes were also reflected among Borbieva’s (2012) respondents, who viewed kidnapping as “just our tradition.” Our findings add to a growing body of research (Becker et al. 2017; Steiner and Becker 2019) that finds that bride kidnapping does have important consequences for families in Kyrgyzstan. Unlike previous research that focuses on how bride kidnapping affects the quality of the marriage, we show that kidnapping may influence the entire household. The migrants identified in this survey are typically the young adult children (or children-in-law) of household heads, demonstrating the possibility of long-term and even multi-generational effects of bride kidnapping. Future research on migration in Kyrgyzstan and other countries where forced marriage is common should consider marriage dynamics, and measurement of the timing of the marriage relative to the timing of migration, and greater detail on the circumstances of the marriage, would address some of the limitations of this study.

Second, our findings support the contention that using labor migration as a coherent, household-level economic strategy works best when the household is hierarchically organized under a strong (and usually male) head (Cerrutti and Gaudio 2010; Massey et al. 2006; Sana and Massey 2005). Although further research is needed to assess whether men and women in kidnap marriages profess a stronger attachment than their peers to traditional or patriarchal gender ideologies, kidnap households in this survey do seem to be more able to tap into labor migration as an economic strategy. Future research, using instrumental variables or other approaches to address endogeneity between household wealth, migration, and remittances, will likely clarify the role of migration in household financial strategies in Kyrgyzstan.

Research in Latin America shows that the migrants sent by these patriarchal households are typically men (Cerrutti and Gaudio 2010; Massey et al. 2006; Sana and Massey 2005). This does not seem to be the case in southern Kyrgyzstan. Strong patriarchal norms do not necessarily serve as a constraint to women’s migration, as some previous studies have found (Cerrutti and Gaudio 2010; Oishi 2005). Instead, they may facilitate both men’s and women’s migration to nearly equal degrees. This finding is consistent with research from other regions of Asia, where young women are seen as ideal migrants because their parents perceive them as more reliable remitters than young men are (Curran et al. 2005; Semyonov and Gorodzeisky 2005). Perhaps even more importantly, Central Asian men in Russia are vulnerable to harassment and violence at the hands of both nationalist groups and local authorities (Khalioullina 2016; Kuznetsova and Round 2019), and this vulnerability is widely understood in Kyrgyzstan (Gerber and Zavisca 2019). Women are vulnerable as well, but they are at least perceived to be treated better in Russian than their male counterparts, making women’s migration feel like a safer option.