Intergenerational Power Shift and the Rise of Nonarranged Marriages Among Refugees

Abstract

The experiences of war and refugee status can alter intrafamily dynamics, with implications for family formation, including marriage. We use the nationally representative Syrian Migrant Sample of the 2018 Turkey Demographic Health Survey (TDHS-S) to conduct a duration analysis of marriage outcomes among Syrian refugees in Turkey, tracking women throughout their residence in prewar Syria (before the conflict began in 2011), postwar Syria (after the conflict began but before arrival in Turkey), and Turkey. We find that early marriage was more prominent among refugees who were unmarried at the time of migration than among those married before migration; the mean marriage age dropped from 19.6 in prewar Syria to 19.1 in postwar Syria and 18.1 in Turkey. Using the TDHS-S and prewar Syrian surveys, we show that this finding aligns with the observed declines in household income and young women's opportunity cost of marriage. Our duration analysis also reveals a notable shift from traditional arranged marriages to more modern marriage forms among refugees in Turkey. An intergenerational power shift might drive the shift toward nonarranged marriages. After arrival in Turkey, wealth and employment of parents decline among refugees. In contrast, Syrian youth in Turkey have higher age-adjusted employment rates than in prewar Syria. Moreover, nonarranged marriages increase more among demographic groups with stronger intergenerational power shifts than among groups with weaker shifts.

Introduction

In 2018, an estimated 70.8 million people worldwide were forcibly displaced as a result of persecution, conflict, violence, or human rights violations. Of these, 25.9 million were refugees, according to the UNHCR (2019). The Syrian civil war has been the largest driver of recent refugee flows, with millions forced to flee their homes since 2011. More than 6 million Syrians are now displaced within Syria, and another 6.7 million have sought refuge in neighboring countries, such as Turkey, Jordan, and Lebanon. Turkey hosts the largest population of Syrian refugees, at almost 3.6 million. As with all refugee populations, Syrians who have fled the civil war and resettled in Turkey have experienced significant changes in their living conditions. Although research has examined the effects of displacement on refugees’ working conditions, health, and education, little attention has focused on how forced migration affects marriage outcomes, despite the frequency of teenage marriage among Syrian refugee women.¹ In this study, we seek to fill this gap by exploring the impact of the Syrian civil war and resulting refugee status on marriage practices among Syrian women living in Turkey.

We also explore the shift in marriage practices among Syrian refugees from arranged marriages to more modern marriage forms, such as “love” marriages. The comparison of arranged versus nonarranged marriages is important because of documented strong associations between nonarranged marriage and women's empowerment indicators, such as asset ownership, decision-making power, freedom from physical violence, and the division of household labor (e.g., for Turkey, see Fox 1975; for India, see Banerji and Deshpande 2021; Jejeebhoy et al. 2013).² Nonarranged marriage was also reported to be positively associated with marital stability and quality of spousal relationship (e.g., for Pakistan, see Hamid et al. 2011; for Indonesia, see Heaton et al. 2001; for Turkey, see Olcay Imamoğlu et al. 2019; for China, see Pimentel 2000; Xu and Whyte 1990) and with women's reduced likelihood of reporting depression (Zhang and Axinn 2021). Consistent with other literature on the subject, our examination of Syrians in Syria during the prewar period, Syrian refugees in Turkey, and the native population in Turkey shows better outcomes for women in nonarranged marriages in all three populations.³

We hypothesize that changes in marriage patterns among Syrian refugees at least partly reflect children's and parents’ changing economic circumstances resulting from their transition to refugee status. In particular, we build on recent work emphasizing the role of economic and social transition in changing marriage-market processes. Corno et al. (2020), Corno and Voena (2023), and Hoogeveen et al. (2011) explored the effect of income shocks on marriage timing in developing countries and how this effect might change by social norms. Baird et al. (2019), Buchmann et al. (2021), and Corno and Voena (2023) examined how cash transfers alter marriage timing.

We draw on data from the Syrian Migrant Sample of the 2018 Turkey Demographic Health Survey (TDHS-S), which is the first survey to include a representative sample of Syrian refugees in Turkey. This survey provides rich information on women's marriage and migration histories and their origins in Syria. We also incorporate information on marriage from household surveys in prewar Syria. Using the retrospective event histories for marriage, we conduct a discrete-time duration analysis. In this structure, different spells of a given woman's life come from distinct periods in prewar Syria (before the conflict began in 2011), postwar Syria (after the conflict began but before arrival in Turkey), and Turkey. Using this data structure, we investigate how refugees’ marriage hazard rates (and marriage-type hazard rates) are impacted by two critical junctures in their lives: (1) the civil war in Syria and (2) refugee status in Turkey. We also provide evidence of the changing economic circumstances of parents and children in refugee households.

Related Literature

The literature suggests that conflict and displacement increase early-marriage rates, particularly among women (Human Rights Watch 2011; Knox 2017; Lu et al. 2021; UNFPA 2018). Previous research has also highlighted refugee children's vulnerability to risks, such as child marriage in Turkey, where the largest number of registered refugees reside (Sahin et al. 2021; Wringe et al. 2019). Some studies found increases in teenage marriages among Syrian refugees, particularly in Lebanon and Jordan (Baird et al. 2022; Bartels et al. 2021; DeJong et al. 2017; Save the Children 2014; Sieverding et al. 2018).⁴ However, these studies have not used population-representative data and thus are not easily generalized.

Koç and Saraç (2021) conducted the sole quantitative study to examine the relationship between conflict and marriage outcomes among Syrian refugees in Turkey. Comparing the marriages realized before and after migration to Turkey among Syrian refugee women in the 2018 TDHS-S, they found a significant decrease in traditional marriage practices among women who married postmigration.⁵ However, as they noted, they compared very different groups: women who chose to marry early in Syria and those who chose to marry late in Turkey. In contrast, our study tracks each woman from an early age until marriage—during prewar Syria, postwar Syria, and while holding refugee status—and estimates the effects of war and refugee status on marriage outcomes. As discussed later, our findings are quite different. We find that both war and refugee status increase traditional marriage hazard rates. However, conditional on marriage, we find a drop in traditional marriage outcomes. In addition, Koç and Saraç (2021) did not examine the underlying mechanisms behind their findings, highlighting a need for research exploring the causes behind these changes.

Our study also examines the potential mechanisms that could lead to a higher probability of women's marriage during the civil war and in refugee status. The literature points to three primary factors in the rise in marriage rates in conflict and displacement settings (Bartels et al. 2021; Elnakib et al. 2021). First, conflicts can decrease households’ economic power because of the destruction of assets and limited access to employment and earnings. In such cases, families may choose to marry off their daughters at a young age as a means of coping with financial difficulties. Second, the disruption of education during conflict and displacement can lower the opportunity cost of marriage for young girls, making it a more attractive option. Finally, conflict and displacement can pose safety concerns for girls, particularly in terms of increased risks of sexual violence and harassment, and families may perceive marriage as a way to protect their daughters. Conversely, as found in a mixed methods study of early marriage among refugees in Lebanon (Bartels et al. 2018), daughters might seek early marriage to escape a lack of freedom at home imposed because of security concerns. However, changes in the economic power of adult children with respect to their parents and falling opportunity costs of marriage might have the same effect.

Despite the prevalence of arranged marriages in various parts of the world, a limited number of studies have focused on parental involvement in decision-making and marriage type in this context (Edlund and Lagerlöf 2006; Fafchamps and Quisumbing 2008; Kedir and Oterová 2017; Mathur 2007; Rosenzweig and Stark 1989; Rubio 2014). Further, little evidence exists about how and why arranged marriages are changing in developing countries, and no study has examined how conflict and forced migration affect these dynamics. Mathur (2007) argued that the choice between arranged and nonarranged marriage results from bargaining between parents and children with different preferences for spouse attributes. The study found that stronger financial and kinship ties between parents and sons increase the likelihood of an arranged marriage. Fafchamps and Quisumbing (2008) examined marriage determinants in developing countries, emphasizing the role of insurance, savings, capital accumulation, and parental involvement. Their findings suggest that parental involvement in children's mate searches increases when assets are transferred to children during marriage. Although these studies highlight the importance of financial ties between parents and sons and transfers to children during marriage in influencing arranged marriages, they did not consider how changes in outside economic opportunities for children and their parents affect marriage decision-making.

Outside the field of economics, a broader literature has examined marital behavior. Drawing on modernization theory, Goode (1963) posited that urbanization and industrialization would lead societies worldwide to evolve toward Western-style companionate marriages that emphasize mutual affection and partnership. Goode argued that this shift toward families in which marital bonds are prioritized over extended familial ties is a natural response to the demands of an industrial economy, which values mobility, education, and workforce flexibility over traditional family structures governed by elder control and arranged marriages. Other research across diverse cultures (Cherlin 2012; Ruggles 2012) revealed a shift among young adults toward prioritizing romantic love and shared interests in their relationships, supporting Goode's prediction of the universal appeal of companionate marriage concepts.

Thornton's developmental idealism theory adds to this discourse by positing that adopting Western family norms is not merely a by-product of economic development but also a result of the widespread belief in the inherent goodness and desirability of Western societal structures, including family dynamics (Thornton 2005). This belief system suggests a causal link between development and adopting the modern family model, which advocates for equality and consent-based relationships. Developmental idealism theory has been supported by findings in Nepal (Allendorf and Thornton 2015), China (Lai and Thornton 2015), and Taiwan (Thornton and Lin 1995) and partially supported by research in India (Allendorf and Pandian 2016; Sarkar and Rizzi 2024). Therefore, although modernization theory emphasizes economic factors and developmental idealism focuses on ideational influences, both theories foresee the decline of arranged marriages (Allendorf and Pandian 2016)—as we find for Syrian refugees in Turkey.

Thornton et al. (2021) identified migration as a key mechanism for spreading developmental idealism, especially through migrants’ interactions with the local population in new environments. Classical assimilation and acculturation theories suggest that such migration experiences profoundly transform migrants’ identities and attitudes (Portes et al. 1980). Migration is often seen as a fast track to modernization, with the rapid influence of social factors linked to development leading to significant changes in migrants’ perceptions and behaviors (Reniers 2001). Through this lens, migrants are expected to have fewer arranged marriages than those in their home countries, indicating a decline in the transmission of traditional marriage customs from one generation to the next among migrant communities (Baykara-Krumme 2017). Additionally, the scarcity of suitable partners in the migrant community might further reduce arranged marriages because fewer families are engaged in maintaining this tradition.

Forced migration disrupts the traditional family structures that support arranged marriages, diminishing the transmission of these practices through generations. As large families are fragmented by migration or loss, the cohesive unit that once decided about marital arrangements is weakened (Tiilikainen et al. 2023), potentially reducing the likelihood of the intergenerational transmission of arranged marriage norms in refugee communities. Supporting this perspective, anecdotal evidence underscores how Syrian refugees, in adapting to new environments, have transformed engagement traditions by utilizing clubs and social media to find marriage partners, showcasing a shift toward modern marriage practices among refugee youth (Khattab 2024).

Although the implications of both the modernization theory and the development idealism theory are consistent with our findings regarding the rise in nonarranged marriages, they are not easily reconciled with the increase in teen marriage often found in refugee populations. It is not obvious that the mechanisms thought to be responsible for early marriage among refugees would encourage the adoption of modern or Western conceptualizations of marriage. For example, decreases in education and wealth should, in principle, slow ideational change, and parents’ security concerns should lead to restrictions on young women's mobility.

In contrast, we emphasize how the economic circumstances of young adults change vis-à-vis their parents, which should increase both nonarranged and early marriages. Parents’ decreased assets and the disruption of their social networks decrease their ability to dictate their children's choices regarding marriage timing and type. Reductions in schooling and work decrease the economic costs of marriage in a setting where marriage is seen as incompatible with these activities. Furthermore, daughters might choose marriage, especially to someone outside of the parental network, to escape the restrictions of parental authority that arise from security concerns.

Background Information

After the March 2011 anti-government protests in Syria spiraled out of control and eventually led to a nationwide civil war, Syrians began seeking refuge in neighboring countries. Since 2011, more than 14 million Syrians have fled their homes for security. More than 6.8 million Syrians are still displaced within their country. Neighboring countries—particularly Turkey, Lebanon, and Jordan—host more than 80% of Syrian refugees. Turkey shares the longest border (911 km) with Syria, and the first mass movement of people from Syria to Turkey occurred on April 29, 2011. The influx of Syrian refugees into Turkey remained intense until 2017. According to Turkey's Presidency of Migration Management (TPMM), 3,644,342 individuals from Syria were under temporary protection as of February 2019, when the 2018 TDHS-S interviews were concluded. This figure represents more than 4% of Turkey's total population of 85 million. Turkey has hosted the most refugees globally since 2014 (UNHCR 2023). By the end of 2019, only 1.7% of Syrians in Turkey lived in temporary accommodation centers or camps, indicating that most Syrian refugees in Turkey have settled as urban refugees.

TPMM (2019) official statistics show that Syrian refugees in Turkey are, on average, younger than the overall population of Turkey. In 2018, the average age of Syrian refugees was 22.5 compared with 31.7 for natives. The youth (aged 15–24) population rate among Syrian refugees, at 22.55%, was higher than Turkey's youth population rate of 15.8%. When broken down by gender, the proportion of males was higher among Syrian refugees than among natives (54% vs. 50%).

Because of the absence of official statistics, we rely on surveys to obtain insights into the education level and employment status of Syrian refugees in Turkey. A comparison of the native and refugee samples of the 2018 TDHS reveals that Syrian refugees possess lower education levels than the national average in Turkey. For individuals aged 18–65, the proportions of Turkish males and females without formal education stand at .02 and .13, respectively; the respective figures for Syrian males and females are .11 and .20. The proportions of Turkish males and females with an education level of high school or above stand at .25 and .22, respectively, while the corresponding figures for Syrian males and females are .09 and .07.

A significant fraction of Syrian refugees are employed. Demirci and Kırdar (2023) reported that among those aged 18–59 in the 2018 TDHS, 61.8% of Syrian refugee men had a paid job compared with 68.9% of native men. In contrast, Syrian women aged 18–59 had a much lower paid employment rate (6.0%) than same-aged native women (22.2%). Similarly, Dayıoğlu et al. (forthcoming) reported high paid employment rates among Syrian male youth: 48.0% of boys aged 15–17 were in paid employment. However, the corresponding rate was much lower for girls of the same age, at 8.5%.

Data and Estimation

The data come from the Syrian Migrant Sample of the 2018 Turkey Demographic and Health Survey, conducted by the Hacettepe University Institute of Population Studies. The TDHS-S provides the first nationally representative, household-level demographic and health indicators for Syrian refugees living in Turkey. The data were collected in November 2018–February 2019 (Hacettepe University Institute of Population Studies 2019) using multistage stratified sampling of households. Among the households available for interview, 1,826 (95%) were successfully interviewed. All women aged 15–49 in the selected households were eligible for the Woman's Questionnaire, regardless of their marital status. Interviews were completed with 2,216 of these women (93%). The Woman's Questionnaire covers rich information on women's demographic and socioeconomic characteristics, including marriage history, marriage characteristics, and husband's background characteristics. The TDHS-S also covers a detailed migration history for all women, including the year of arrival in Turkey. In addition, we use two complementary datasets—the 2009 Syria Family Health Survey (SFHS) and the 2006 Syria Multiple Indicators Survey (SMICS)—for background information on Syrian women before they arrived in Turkey (PARFAM 2009; World Bank Group 2022).

We examine seven outcome variables from the TDHS-S: (1) married refers to women's marital status; (2) arranged marriage refers to traditional, family-framed encounters; (3) nonarranged marriage pertains to marriages the couple arranged without parental intervention; (4) husband paid bride price indicates that the groom or his family paid the bride-price in cash/gold or in kind; (5) only religious ceremony indicates that the couple did not have a civil marriage ceremony; (6) husband has other wives refers to polygamous marriages; and (7) woman is related to her husband pertains to first-cousin marriages. The control variables include age and a set of individual characteristics not affected by marriage: birth year, mother's and father's education levels, mother tongue, type of birthplace (province center, district center, or subdistrict/village), and birth province (covering 14 Syrian provinces).

Using the TDHS-S, we construct retrospective event histories for marriage outcomes. In particular, we format the data for a discrete-time duration analysis in which each period is one age and marriage constitutes the event of interest. The event history for all women starts at age 10, the youngest age at marriage in the data. The event history continues until the age at first marriage for ever-married women and until the age at the survey year (2018) for never-married women. For ever-married women, the outcome variable takes the value of 1 at the age at marriage and 0 at all other ages. For never-married women, the outcome variable is right-censored and takes the value of 0 at all age values. When we structure the data into person-age observations, we obtain 21,572 observations for 2,212 women. We aim to measure the impact of two critical junctures in refugee women's life cycle: (1) the onset of the Syrian civil war in 2011 and (2) their arrival in Turkey and the beginning of their refugee experience. For this purpose, we generate a location indicator variable that takes one of three values: (1) the prewar Syrian period, (2) the postwar Syrian period (i.e., after the civil war began), and (3) the period in Turkey as a refugee. Table A2 (online appendix) provides an illustration of the data structure.

Using this data structure, we estimate the following specification:

In Eq. (1), $mit$ takes the value of 1 if woman i (who is never-married) marries at time t and 0 otherwise; $(postwarSyria)it$ takes the value of 1 for woman i after 2011 but before her arrival in Turkey and 0 otherwise; and $(inTurkey)it$ takes the value 1 for woman i after her arrival in Turkey and 0 otherwise. We control for age using a dummy variable for each age. Finally, X represents the control variables. The key coefficients of interest, $β1$ and $β2$⁠, show the difference in the marriage probabilities between postwar Syria and prewar Syria and between Turkey and prewar Syria, respectively. We estimate Eq. (1) using a linear probability model (LPM), but we use a logit regression to check the robustness of the estimates.

Descriptive Statistics

Table A3 (online appendix) presents the mean values of the nine marriage and birth outcomes listed earlier, as well as age. The first column presents the values for the total sample. The second, third, and fourth columns give the values for the three phases Syrian refugees experienced: prewar Syria, postwar Syria, and Turkey. An analysis of the temporal variation of dependent variables reveals an increased frequency of total marriages, arranged marriages, and nonarranged marriages in Turkey relative to Syria. However, this increase is largely attributable to the older age of refugee women in Turkey, which is accounted for in the estimation through the inclusion of age effects.

According to Table A3, of all the woman-age observations in the sample, marriage is observed in 8.6% of the observations. Arranged marriages are more common than nonarranged marriages, occurring in 6.8% versus 1.8% of the observations, respectively. The payment of a bride-price and first-cousin marriages are also frequently observed, occurring in 6.3% and 4.0% of the observations, respectively.

Figure 1 illustrates the marriage hazard rates for prewar Syria, postwar Syria, and Turkey. Panel a displays rates for all marriage types, whereas panels b and c provide rates for arranged and nonarranged marriages, respectively. Three observations are evident from a comparison of these panels. First, in all three periods, most marriages were arranged. Second, for all marriage types, the hazard rates during the postwar Syria period and in Turkey are higher than those during the prewar Syria period. Moreover, overall, the hazard rates in Turkey are somewhat higher than those in postwar Syria. Third, the proportional increase in the hazard rates during the postwar Syria period and in Turkey relative to those in the prewar Syria period is particularly large for nonarranged marriages.

Results

Main Results

Table 1 presents the results of estimating Eq. (1) by an LPM. The analysis primarily concentrates on the influence of the in Turkey and postwar Syria dummy variables, with the omitted category being prewar Syria. We also present the difference between the two key variables of interest to understand the change in Turkey relative to postwar Syria.

First, we explore the changes in marriage patterns with the onset of the war. The marriage hazard rate increases by 3.4 percentage points (pp) with the war onset. A larger fraction of this rise comes from the increase in the hazard rate of nonarranged marriages, although the estimates also indicate a statistically significant rise in arranged marriages. We also observe increases in the hazard rates of marriages in which the husband paid bride-price and the woman is related to her husband at a level comparable to the rise in arranged marriages. In addition, we observe a significant increase in the hazard rate of marriages with only a religious ceremony, which is higher than the rise in total marriages, suggesting a shift from both civil and religious ceremonies to only religious ones.

Next, we compare marriage patterns in prewar Syria and Turkey. The marriage hazard rate is 6.6 pp higher in Turkey than in prewar Syria. The difference between the two key variables indicates a significant increase in the marriage hazard rate (by 3.2 pp) after Syrians arrive in Turkey from postwar Syria. Focusing on marriage types, we observe that the hazard rates for arranged marriage and nonarranged marriage in Turkey are 3.0 pp and 3.6 pp higher than the corresponding levels in prewar Syria. However, the mean values shown at the bottom of Table 1 (the marriage hazard rates averaged across all ages) indicate that the percentage increase in the hazard rate for nonarranged marriage is significantly higher than that for arranged marriage, as further discussed in the next section. Furthermore, both arranged and nonarranged marriage hazard rates rise as refugees arrive in Turkey.

Table 1 also shows that the hazard rate of marriages involving bride-price is 2.6 pp higher in Turkey than in prewar Syria, which can be attributed to the rise of arranged marriages (by 3.0 pp). This traditional practice is prevalent in marriages arranged by families. In addition, column 5 indicates that the hazard rate of marriages conducted solely through a religious ceremony is higher in Turkey than in prewar and postwar Syria. The rise in Turkey might be surprising at first, given that civil marriage is technically mandatory. However, religious marriages occur among the native population in Turkey, and sole religious marriages are likely at early-age marriages (particularly before the mandated minimum age of marriage). More importantly, civil and religious marriages take place at separate places in Turkey, whereas they can occur together in Syria. Finally, we find higher first-cousin marriage hazard rates in Turkey than in prewar or postwar Syria—parallel to the rise in arranged marriages. However, the magnitude of this increase (1.8 pp) is smaller than that for arranged marriages (3.0 pp).

In an alternative specification, we extend Eq. (1) by interacting the in Turkey dummy variable with the years of residence in Turkey to examine variation in the change in the marriage hazard rate in Turkey over time. The results in Table A4 (online appendix) show that the arranged marriage hazard rate is 5.0 pp higher during the first year in Turkey than in prewar Syria. However, the rise in this hazard rate diminishes over time in Turkey, such that almost no difference exists between the hazard rates in prewar Syria and Turkey after four years in Turkey. In contrast, the difference between the nonarranged marriage hazard rate in Turkey and prewar Syria does not decrease over time; in fact, the estimated coefficient of the interaction term is virtually zero. Finally, as a robustness check, we estimate Eq. (1) using a logit regression instead of an LPM. The results in Table A5 (online appendix) show that all findings persist, except that the impacts of postwar Syria and Turkey dummy variables on arranged marriage become marginally statistically insignificant.

Effects on Age-Specific Hazard Rates

The preceding section presented findings obtained from aggregating data across all age groups. This section examines the effects of being in Turkey or postwar Syria on the marriage hazard rate by age. The first column of Figure 2 compares Turkey with prewar Syria, and the second column compares postwar and prewar Syria. Panel a displays statistical evidence that the marriage hazard rate in Turkey is higher than that in prewar Syria for ages 17–24. At these ages, the marriage hazard rate in Turkey is almost 10 pp higher. Panel b on arranged marriages shows that statistical evidence of a higher hazard rate in Turkey than in prewar Syria exists only for age 17, and the gaps are marginally statistically insignificant at the 10% level for ages 18 and 19. In contrast, panel c on nonarranged marriages displays statistical evidence of a gap at ages 18–22, with marginally statistically insignificant gaps at the 10% level at ages 23 and 24. The timing of the effects on arranged and nonarranged marriages is consistent with Figure 1, which illustrates a peak in the arranged marriage hazard rate at age 17. In contrast, nonarranged marriages peak in the early 20s. Finally, using the age-specific baseline hazard rates given in Figure 1, we calculate the percentage change in the hazard rates that the estimates coefficients in panels b and c imply. We find that for refugees aged 15–25, the percentage increase in the hazard rate is roughly 100% for nonarranged marriages but less than 50% for arranged marriages.

The second column of Figure 2 demonstrates changes in the age-specific marriage hazard rates in postwar Syria. The figure provides statistical evidence, at least at the 10% level, that war increases hazard rates during the late teenage years (ages 17–19).⁶ In panel b, the patterns for arranged marriages in the second column are similar to those in the first column; however, the increase in hazard rates during the late teenage years is, overall, less precisely estimated. Panel c, however, suggests that the hazard rates of nonarranged marriages are higher in the late teenage years and the early 20s. As in the impact of refugee status in panel a, the impact of the civil war is realized, on average, at higher ages for nonarranged marriages than for arranged marriages.

Finally, using the predicted age-specific hazard rates in prewar Syria, postwar Syria, and Turkey, we calculate the predicted mean age at marriage in each period. We find a mean age at marriage of 19.6 in prewar Syria, which drops to 19.1 in postwar Syria and 18.1 in Turkey.

Understanding the Rise in Marriage Rates

This section explores the factors driving the observed rise in marriage hazard rates, including the impact of increasing poverty among refugee families and the role of the opportunity cost of marriage for young women in terms of schooling and employment rates. Poverty can drive increased marriage rates in the context of armed conflict and forced displacement. Such situations are often marked by a loss of employment and economic opportunities, causing financial distress for many families and potentially leading parents to marry off their daughters earlier to reduce the household's consumption needs. To investigate whether deteriorating economic conditions could be driving the observed trends, we compare both the employment outcomes of male Syrian refugees and asset holdings of Syrian households in Turkey with those in prewar Syria. Here, the prewar Syria data on employment outcomes come from the 2009 SFHS, and the prewar Syria data for asset holdings come from the 2009 SFHS and 2006 SMICS. Because Syrian refugees more likely originate from northern Syria and regional differences are important in Syria, we weight the province-specific averages by the fraction of Syrian refugees in the 2018 TDHS-S who originated from each province.

Table 2 compares Turkey and prewar Syria in employment outcomes (panel A) and asset holdings (panel B). Employment and wage employment are available in the 2009 SFHS, but the 2018 TDHS-S measured employment only for wage workers. However, the 2018 TDHS-S assessed both employment and wage employment for married men and women (because more detailed questions are elicited about partners from the target female population). We impute employment rates in Turkey using Syrian men's wage employment rates in Turkey and both employment and wage employment rates of married Syrian men in Turkey.⁷

Panel A of Table 2 shows that Syrian men's employment rate is much lower in Turkey than in prewar Syria (67.1% vs. 93.2% for married men; 69.9% vs. 81.8% for all men). The wage employment rates of married men and all men in Turkey are similar, and the wage employment rate of all men is somewhat higher in Turkey than in prewar Syria (61.8% vs. 52.2%) because refugee men's nonwage employment rate is much lower (Demirci and Kırdar 2023). Among Syrian women, both employment and wage employment rates are lower in Turkey than in prewar Syria. The employment rate in Turkey is about half that in prewar Syria. These patterns suggest that refugees’ household labor income is significantly lower in Turkey than in prewar Syria.

Panel B of Table 2 shows that asset ownership is much lower for Syrians in Turkey than in prewar Syria. For instance, among Syrian refugee households in Turkey, 0.5% own a house and 2.9% own a car, compared with 77.9% and 16.0%, respectively, for households in Syria. In addition, the ownership rate of housing items, such as washing machines, dishwashers, and computers, is lower in Turkey than in prewar Syria. Therefore, considering the decrease in employment rates and asset holdings, poverty might be a valid explanation for the increased marriage rates among Syrian refugees.

To better understand Syrian refugees’ wealth status, we generate a wealth index and compare the native Turkish and Syrian refugee populations on this measure. In constructing the wealth index, we consider household assets, and each item a household owns contributes 1 point to the wealth index.⁸Figure A1 (online appendix) highlights a stark contrast between the wealth distributions, revealing that approximately 44.7% of Syrian refugee households are clustered in the lowest decile, followed by 35.2% in the second-lowest decile and 15.4% in the third-lowest decile. In contrast, only 1.8% of native Turkish households are in the lowest decile, 5.8% are in the second-lowest decile, and 11.7% are in the third-lowest decile. In other words, the total percentage of households in the bottom three deciles is below 20% for Turkish natives but roughly 95% for Syrian refugees. In essence, the figure shows significantly lower levels of wealth for Syrian households than for their Turkish counterparts.⁹

Another reason for the rise in the marriage hazard rates could be a fall in the opportunity cost of marriage, which is the value of school enrollment and employment. Hence, we investigate how young women's school enrollment and employment rates compare in prewar Syria and in Turkey among refugees. Panel A of Table 3 shows that female enrollment rates are much lower in Turkey, and the difference is particularly striking for the 15–19 age group. The female enrollment rate was 40.0% in prewar Syria compared with 15.1% in Turkey at ages 15–19 and 14.3% in prewar Syria compared with only 3.5% in Turkey at ages 20–24. Similarly, employment rates are lower in Turkey than in prewar Syria for young women, although the gap in wage employment rates is smaller. Among women aged 20–24, 14.0% were employed in prewar Syria compared with 6.0% in Turkey; among women aged 25–29, 17.8% were employed in prewar Syria compared with 8.5% in Turkey.

The evidence provided in this section reveals that the observed rise in marriage hazard rates can be attributed to two primary factors: (1) an increase in poverty among refugee families, as evidenced by declining employment rates of family members and decreases in family assets; and (2) a decrease in the opportunity cost of marriage for young women in terms of their education and employment opportunities.

Understanding the Larger Increase in Nonarranged Marriages

This section considers the significant rise in nonarranged marriages among Turkey's Syrian refugee community. The first potential explanation for this change is an intergenerational power shift in which younger generations gain greater economic independence, becoming less dependent on their parents for mate selection and marriage expenses. A second potential factor is the higher returns to mobility in Turkey, giving refugees greater access to higher-paying jobs elsewhere by making more independent decisions about their marriages.

Intergenerational Power Shift

One previously explored potential explanation for choosing between arranged and nonarranged marriages is bargaining between parents and children. Arranged marriages are characterized by a higher degree of parental control over the selection of spouses and are more likely when financial ties between parents and sons are stronger (Mathur 2007). Parents tend to be more involved in their children's mate search when assets are transferred to children during the marriage (Fafchamps and Quisumbing 2008) and when parents have more power than their children (Edlund and Lagerlöf 2004). However, as younger generations gain greater economic independence and employment opportunities, they become less dependent on their parents for mate selection and marriage expenses, which might lead to a transition toward nonarranged marriages. Therefore, an intergenerational power shift from parents to children may play a role in shaping the prevalence of nonarranged marriages.¹⁰

As discussed earlier, households’ asset holdings were significantly lower in Turkey than in prewar Syria (see Table 2), suggesting lower bargaining power for parents in their children's marriage decisions. Another potential source of an intergenerational power shift is an age-specific difference in refugees’ employment outcomes in Turkey and prewar Syria. Table 4 displays men's age-specific employment and wage employment rates in prewar Syria and Turkey.¹¹ We find that although younger male workers (aged 15–24) were more likely to be employed in Turkey, the employment rate was higher in prewar Syria for all other age groups. The gap becomes substantial (35 pp or more) after age 45. The primary reason for this difference by age is that most employed Syrian refugees in Turkey are wage workers in the informal sector. Because most of these jobs require physical power, age is a critical determinant of refugees’ employment status (Demirci and Kırdar 2023). The gap in wage employment between prewar Syria and Turkey, as shown in Table 4, is less acute because of the higher likelihood of wage employment among employed Syrian men in Turkey.¹² Nonetheless, even for wage employment, the fraction is significantly higher in prewar Syria for ages above 45.

Comparing employment rates among different age groups indicates that older males experienced significant losses in employment opportunities after migrating to Turkey, whereas younger males were comparatively better off. This shift in employment opportunities suggests a potential intergenerational power shift, with younger generations gaining more control over their economic prospects.

To further assess the role of an intergenerational power shift, we exploit differences in age-specific changes in economic opportunities across demographic groups. We expect that groups experiencing a greater shift in power will exhibit a more pronounced increase in nonarranged marriages. To test this hypothesis, we divide the male sample into those with less than a high school education and those with at least a high school education and compare their employment rates. We focus on older males to examine potential fathers. As shown in panel A of Table 5, the wage employment rates for low-educated men aged 45–49 and 50–54 were, respectively, 0.53 and 0.43 in Syria compared with 0.37 and 0.18 in Turkey. Thus, less educated Syrian men experienced significant employment losses in Turkey. Conversely, high-educated men aged 45–49 and 50–54 exhibited similar wage employment rates in Syria and Turkey: respectively, 0.60 and 0.61 in Syria and 0.59 and 0.41 in Turkey. This finding indicates that more educated Syrian men experience no employment losses. Thus, the intergenerational power shift is more significant for older men (who are likely to be fathers) who are less educated.

If the rise in nonarranged marriages results from an intergenerational power transfer, nonarranged marriages will increase more among children of men with lower educational attainment, who experience a greater shift in power. To investigate this hypothesis, we divide our sample based on the father's educational attainment and examine the effect on marriage outcomes. The estimation results in panel B of Table 5 show that the increase in nonarranged marriages is high and precisely estimated for women with low-educated fathers. In contrast, the effect on nonarranged marriages is small and statistically insignificant for women with high-educated fathers. This finding indicates that the rise in nonarranged marriages is more pronounced for the group with a greater intergenerational power shift: the low-educated group.

Like the analysis in Table 5 by fathers’ educational attainment, Table 6 examines the differences between refugees with rural and urban backgrounds in age-specific changes in economic opportunities and marriage patterns. Almost all Syrian refugees in Turkey live in urban areas. However, before the war, only 62% of Syrian refugees in our sample lived in urban areas of Syria (in provinces or district centers). Panel A of Table 6 shows that among potential fathers (men aged 30 or older), the fall in employment rates after arrival in Turkey was higher for those with rural backgrounds. This finding holds for all age groups, and the gap is large for the age groups most likely to include the fathers of marriage-age women. For instance, the drops for urban men and rural men are, respectively, 7% and 20% for ages 40–44, 33% and 41% for ages 45–49, and 50% and 69% for ages 50–54. The changes in wage employment rates are also more adverse for men of rural origin, regardless of the age group. In addition, using the 2009 SFHS, we calculate that the house ownership rates were 72.4% for rural households and 82.7% for urban households. These rates suggest a larger wealth loss in terms of housing for rural households, given that the house ownership rate is below 1% for rural- or urban-origin refugees. Therefore, panel A suggests a stronger intergenerational power shift in households of rural origin.

The finding in panel A implies a larger increase in nonarranged marriages in rural areas. Panel B of Table 6 shows that the marriage hazard rate is 6.6 pp higher in Turkey than in prewar Syria for women with rural and urban origins in Syria. Examining the impacts by marriage type, we observe that both arranged and nonarranged marriage hazard rates are 3.3 pp higher in Turkey than in prewar Syria for women with urban childhood residence. In contrast, for women with a rural childhood residence, the rise in the nonarranged marriage hazard rate in Turkey (4.1 pp) is higher than the rise in the arranged marriage hazard rate (2.5 pp). Because the baseline level of nonarranged marriages is lower for women of rural origin, their increase in nonarranged marriages is higher both in percentage points (4.1 vs. 3.3 pp) and percentages. Hence, as expected, the rise in nonarranged marriages is higher for the group—with rural background—for whom the intergenerational shift is stronger.¹³

Finally, we try to link women's employment status (as a measure of their intergenerational power) directly to marital outcomes. Using the information on women's employment history after age 12 in the data, we generate a dummy variable for women's employment status at each age and include this variable as an additional control in estimating Eq. (1). We hypothesize that employment status will reduce marriage of either type if marriage is seen as incompatible with women's employment. However, women's employment will result in a shift from arranged to nonarranged marriage if it shifts power from parents to their children. Table A6 (online appendix) shows that women's employment status was associated with a lower propensity for marriage and arranged marriage in Turkey. The estimate for nonarranged marriage is not significantly different from zero. These results are thus consistent with both of our proposed mechanisms. However, we cannot attribute a causal interpretation to the employment status variable here because employment and marriage outcomes are jointly determined. Hence, we provide this analysis only as supportive evidence.

Higher Returns to Migration

An alternative explanation for the larger increase in nonarranged marriages among Syrian refugees is that the returns to geographic mobility are higher in Turkey than in Syria. When the returns to moving to a different location and finding a partner with higher earnings potential are higher, the probability of an arranged marriage decreases, at least to the extent that arranged marriage requires local knowledge of alternative spouses. To examine this potential mechanism, we check whether refugees were more mobile in Turkey than in Syria. We construct retrospective event histories of mobility using migration history since age 10 in the TDHS-S. Specifically, we structure the data into person-age observations from age 10 to the age at the survey. The dependent variable (migration) takes the value of 1 if the woman changes her province of residence at that age and 0 otherwise. This data structure differs from that for marriage because migration can occur several times for the same woman (in Syria or Turkey). We exclude the move from Syria to Turkey because we are interested in the mobility change after refugees arrived in Turkey. The resulting sample includes 2,212 women in 42,652 woman-age observations, and the probability of migration at a given age is 2.1%. We estimate the impact of civil war and refugee status on the likelihood of migrating to another province using an ordinary least-squares (OLS) regression of Eq. (1).

The estimation results in Table A7 (online appendix) show that the probability of migration to another province at a given age is 3.1 pp higher in postwar Syria than in prewar Syria and 3.6 pp higher in Turkey than in prewar Syria. (Both coefficients are statistically significant at the 1% level.) The higher propensity to migrate within Syria after the war is presumably expected as locals flee conflict areas. The elevated likelihood of migrating within Turkey after arrival suggests that refugees are more likely to seek new labor market opportunities by moving to other cities in Turkey. However, arranged marriage could force them to stay in one place and forgo potential returns. Thus, the higher returns to mobility in Turkey might be another reason for the larger increase in nonarranged marriages among Syrian refugees.

Conclusion

In this study, we use nationally representative microdata on Syrian refugees in Turkey (2018 TDHS-S) to examine how two critical junctions in refugees’ life cycle—the onset of the civil war and their arrival in Turkey—impacted women's marriage outcomes. We find that war and forced migration increased the marriage hazard rate among Syrian women. The marriage hazard rate, on average over ages, was 3.4 pp higher in postwar Syria and 6.6 pp higher in Turkey than in prewar Syria. The civil war and refugee status increased both arranged and nonarranged marriage hazard rates; however, the rise in nonarranged marriages was higher in percentage points and percentage terms. The rise in arranged marriages after arrival in Turkey was highly duration-dependent; the increase vanished after a few years in Turkey. In contrast, the rise in nonarranged marriages persisted throughout their residence in Turkey.

Age-specific hazard rates were higher in Turkey than in prewar Syria for ages 17–24. Although the rise in arranged marriages contributed to this increase only in the teenage years, the increase in nonarranged marriage hazard rates contributed in both the late teenage years and the early twenties. Among refugees, the impact of the civil war was more evident at higher ages for nonarranged marriages than for arranged marriages. Using the predicted age-specific hazard rates in prewar Syria, postwar Syria, and Turkey, we calculate that the mean age at marriage dropped from 19.6 in prewar Syria to 19.1 in postwar Syria and 18.1 in Turkey.

Using both prewar data from Syria and the TDHS-S, we identify likely mechanisms for the rise in marriage hazard rates: (1) the deterioration in household economic conditions after forced migration that results from the loss of asset holdings and falling employment rates; and (2) the fall in the opportunity cost of marriage due to lower school enrollment and employment rates of young refugee women in Turkey.

A key finding of this study is that conditional on marriage, Syrian women are more likely to form nonarranged marriages under refugee status. Examining the potential mechanisms for this finding, we find a significant intergenerational power shift after forced migration in which the older generation effectively lost influence in the family. First, the older generation had lower asset holdings after migration that they could transfer to the younger generation. Second, they had much lower employment rates in Turkey, whereas the employment rates of younger refugees did not change much after migration. Consistent with this interpretation, the likelihood of nonarranged marriages rises more among the children of subpopulations who lose more in the labor market (i.e., less educated refugee parents and refugee parents with rural origins).

Examining other potential explanations for the larger increase in nonarranged marriages, we observe that refugees were more likely to migrate interprovincially in Turkey than in prewar Syria. This finding suggests that another reason for the larger rise in nonarranged marriages might be higher returns to mobility due to better labor market opportunities in other regions. Arranged marriages require parental familiarity and access to effective social networks with knowledge about possible partners and are thus likely to inhibit young adults’ migration.

This study significantly contributes to the literature on refugee communities by comprehensively analyzing the changes in marriage practices among Syrian refugees in Turkey, the largest refugee group in a single country. Our approach unveils the underlying reasons for the observed shifts in marriage patterns, such as the rise in marriage rates and the intergenerational power transitions that influence the prevalence of nonarranged marriages. These findings offer valuable insights for future research, illuminating the complex interactions among forced migration, sociocultural transformations, and demographic outcomes. By building on our findings, researchers can further explore the long-term consequences of war and displacement on various aspects of refugee lives, such as family dynamics, gender roles, and social integration. This deeper understanding will help inform evidence-based policies and interventions aimed at addressing the multifaceted challenges faced by displaced populations and promoting their long-term well-being and integration into host societies.

Acknowledgments

We thank Sara Curran, Demography editor in chief; two anonymous referees; and İsmet Koç for several valuable comments and suggestions. We also thank the seminar participants at Sabancı University and workshop participants of MSCA-IF-2020 Global Fellowship, Project 101024877, at Boğaziçi University. G. Kırdar gratefully acknowledges financial support from the European Commission, MSCA-IF-2020 Global Fellowship, Project 101024877. Research reported in this publication was supported in part by the Population Studies and Training Center at Brown University through the generosity of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2C HD041020). The authors are solely responsible for the content expressed herein.

Notes

References