Abstract
This study investigates the difference in rates of sexual assault between left-behind children and those living with both parents in rural China and attempts to identify potential social mechanisms explaining this variation. Using data from a probability sample of middle school students in Guizhou Province, China, our study reveals that parental migration, particularly maternal and both-parent migration, significantly increases children’s risk of sexual victimization. This relationship is mediated by three intervening pathways: weakened caretaker monitoring and supervision, children’s increased engagement in risky lifestyles, and elevated exposure to general victimization. These findings highlight the urgency to develop prevention and treatment programs based on a holistic understanding of protective and risk factors for sexual abuse against left-behind children in rural China.
Sexual abuse against children is a culturally and historically taboo topic in rural China but has gained much public and academic attention in recent years. Whereas the rate of childhood sexual victimization in China is relatively lower than that in western countries (Finkelhor et al., 2013), roughly one-tenth of Chinese children are exposed to sexual victimization, with estimates varying according to the definition of childhood sexual abuse, the nature of sample, and geographic locations (Finkelhor et al., 2013; Luo et al., 2008; Ma, 2018; Stoltenborgh et al., 2011). Numerous studies demonstrate that childhood sexual victimization leads to a variety of adverse developmental outcomes, including depression and anxiety, academic problems, suicidal ideation, and self-harm (Hailes et al., 2019; Jespersen et al., 2009; Luo et al., 2008). The effects of childhood sexual victimization can be further exacerbated among socioeconomically disadvantaged groups, due to their disproportionally higher prevalence of victimization but considerably more limited resources to cope with such a traumatic experience (Assink et al., 2019).
Left-behind children (LBC) in rural China are potentially such a group that is highly vulnerable to sexual abuse. LBC are defined as those who live with one parent, grandparents, or other adults in rural areas when one or both parents migrate to more developed regions for better employment opportunities. Recent national statistics show that the number of LBC reached 61 million in 2011, constituting 38% of the child population in rural China (All-China Women’s Federation [ACWF], 2013). Children left behind, due to insufficient or improper parental or caretaker guardianship, are potentially suitable targets for sexual exploitation in the eyes of offenders. Indeed, one of the major concerns of rural-to-urban migrant parents is their left behind children’s safety and victimization risk (Chen et al., 2017; Toyota et al., 2007), periodically amplified by high-profile reports of severe physical and sexual abuse against LBC. In August 2013, for example, a report that a schoolteacher sexually molested six primary school girls in Jiangxi province, with all of them being LBC, spurred a national debate regarding the social costs of China’s massive rural-to-urban migration (Lau, 2013).
Despite the often dramatized but ovesimplifed news reports, there is little systematic research investigating whether parental migration increases LBC’s odds of sexual victimization in rural China. Whereas anectodal evidence from court records suggests that children left behind are more vulnerable to sexual victimization (Jiang, 2012; Wang et al., 2020; Zhang, 2019; Zhang & Gen, 2016), empirical research systematically investigating this important topic, guided by coherent theoretical frameworks, is scarce. Indeed, to date, we have only found two studies that collected primary self-report data to identify associated risk and protective factors (Chen et al., 2020; Yan et al., 2018). Both are hampered by serious data limitations and produce mostly descriptive and inconsistent findings. Consequently, there is no clear consensus on the prevalence of sexual abuse among LBC, and we know even less about the social processes underlying the connection between parental migration and sexual assault against children left behind.
This research contributes to the limited literature by investigating two fundamental questions. First, does the rate of sexual assault differ between children left behind and children living with both parents? That is, does parental migration increase children’s likelihood of sexual victimization in rural China? Second, if there is such a relationship, what are the possible social mechanisms underlying this relationship? Our study, built on the transitional model of victimization (Belsky, 1980; Cicchetti et al., 2000) that integrates elements from the lifestyle/routine activities theories (Bunch et al., 2015; Cohen & Felson, 1979) and Finkelhor and Asdigian’s revised theory (1996), used a probability sample of middle school students in rural China to investigate these critical questions. Our results provide the much warranted empirical evidence for the dynamic connections between parental migration and LBC victimization through various pathways, based on which insights into priority entry points for prevention and treatment programs addressing sexual abuse against left-behind children in rural China can be generated.
Theoretical Background: Parental Migration and Sexual Assault
Past research on sexual assault is dominated by the lifestyle-routine activities theories, which propose that victimization is non-randomly distributed across social groups but is associated with individuals’ demographic characteristics, lifestyles, and routine activities. Criminal victimization occurs when three factors, motivated offenders, suitable targets, and an absence of capable guardianship, converge in time and space (Bunch et al., 2015; Cohen & Felson, 1979). A person’s vulnerability to victimization thus increases when she/he is close to potential offenders, perceived by these offenders as an attractive target, and lacks capable guardians or place managers in the proximity (Welsh et al., 2015). Primarily applied to property and violent victimization, this theory is later revised and extended by Finkelhor and Asdigian (1996) to accommodate specific victimization types such as sexual assault, which is often committed by intimates such as family members or acquaintances. Acknowledging that many of the risk factors associated with sexual victimization (e.g., family stress, parental fighting, being female or male, low self-esteem, and poor health) can only “to a limited extent be subsumed into the routine activities conceptual framework” (p.5), Finkelhor and Asdigian (1996) propose that “concepts like guardianship, exposure, and proximity. . .need to be seen not as aspects of routine activities or lifestyles, but as environmental factors that expose or protect victims from victimization” (p.6). In other words, risk factors such as parental fighting or a lack of parental supervision are environmental factors that put children into increased risk and are not problems of children’s own routine activities or lifestyles.
Whereas the lifestyle-routine activities theories and its revised version have received much empirical support, some criticize that these theories are static and fail to theoretically accommodate the dynamic interplay among individual, familial, and community risk and protective factors (Belsky, 1980; Cicchetti et al., 2000). Cicchetti and colleagues (2000), for instance, contend that the risk for child abuse is influenced by a complex interaction of risk and protective factors at different ecological levels. Adopting an ecological perspective (Bronfenbrenner, 1979), they argue that these risk and protective factors operate at the macro-level (e.g., societal attitude toward child abuse), the exosystem level (e.g., community and social support), and the microsystem level (e.g., characteristics of children and their families). The interplay of these factors determines which ones directly influence children’s likelihood of victimization or play a more distant and indirect role, and thus gives rise to specific pathways through which the likelihood of children’s victimization is elevated or reduced (Belsky, 1980). In other words, integrating the lifestyle/routine activities theories and their extension, this approach highlights the complex interplay of various factors at different levels and the transactional and dynamic nature of the social processes leading to child abuse.
Our theoretical model is built upon the transactional model of victimization. More specifically, we contend that parental migration in rural China, directly and indirectly, structures and shapes children’s dynamic interaction with peers, school, and at-home caretakers, consequently elevating or reducing their exposure to opportunities conducive to sexual assault. First, we argue that parental migration directly increases children’s vulnerability by creating a prolonged physical—and possibly emotional—separation between children and their migrant parents, which significantly compromises migrant and at-home parents’ and grandparents’ abilities to effectively guard their children against victimization. The long duration of temporal and geographic separation, typical for inter-province and even intra-province migration in rural China (Wen & Lin, 2012), depletes parents’ abilities to protect their children from victimization. Overall, we hypothesize that compared with children living with both parents in rural China, LBC cared by one at-home-parent or grandparents have a higher likelihood of being sexually assaulted.
Second, we propose that, in a transactional manner, parental migration leads to significant changes in children’s relationships with family members, school, and peers, creating unique pathways through which the odds of sexual victimization are decreased or elevated. In other words, we posit that the effect of parental migration on children’s victimization can also be indirect, mediated by a series of intervening environmental factors. For example, among one-parent migrant households, the migration of one parent may significantly increase the at-home parent’s household responsibilities and reduces his/her time spent with children (Chang et al., 2011), leading to children’s increased time with peers or other potential offenders. Additionally, parental migration leads to dramatic change in social roles for aged grandparents, who now become primary caretakers but often lack a basic understanding of sexual abuse or adequate physical and mental strength to protect their grandchildren from potential harm and victimization (Jingzhong & Lu, 2011; Kong & Meng, 2010). Based on the literature (Assink et al., 2019; Vicary et al., 1995), we focus on three sets of potential mediators: children’s bonding with primary caretakers and schools, involvement in risky lifestyles, and concurrent victimization experiences.
Children’s bonding with primary caretakers and schools can serve as the first set of mediators. There is strong evidence that parental migration, particularly maternal migration, weakens LBC’s bonding with primary caretakers (Chen et al., 2015, 2020; Oliveira, 2018; Parreñas, 2005a; Wen & Lin, 2012), although the relationship between parental migration and school bonding is less clear. The recent China Family Panel Studies (CFPS) project, for instance, found that compared with non-migrant parents, caretakers in migrant families interacted less frequently with children left behind, particularly in the area of learning and education (Chen et al., 2015). As a result, LBC, in general, reported a lower level of affectual connection and a higher level of tension and conflict with primary caregivers (Oliveira, 2018; Ye & Murray, 2005). The association between parental migration and children’s school bonding is more equivocal, with some reporting a positive association whereas others finding a null or even negative relationship (Lu, 2012; Sun et al., 2015; Wen & Lin, 2012). Strong bonding with conventional institutions reduces the probability of victimization directly through children’s increased time with family members, teachers, and prosocial peers (Schreck et al., 2002; Tillyer et al., 2016; Zavala et al., 2019). A recent meta-analytic review on childhood sexual victimization provides strong support for this finding, demonstrating that parent-related factors, such as a lack of parental monitoring and care, have small to medium effects on children’s sexual victimization (Assink et al., 2019). These effects can also be indirect, as commitment and attachment to caretakers and school lead to fewer deviant peers and a lower level of delinquency (Sabatine et al., 2017; Shen & Zhong, 2018), which, ultimately, decreases exposure to sexual victimization.
Children’s risky lifestyles, including their association with deviant peers, unsupervised peer activities, and engagement in delinquency, are the second set of intervening variables. An emerging body of research shows that parental migration, in general, is associated with children’s elevated levels of risky and delinquent activities (Chen, 2017; Chen & Jiang, 2019; Gao et al., 2010). Parental migration may indirectly increase children’s involvement in risky lifestyles through a lack of parental involvement, monitoring, and supervision, or directly increase children’s involvement due to a need for emotional support and companionship from similarly situated peers. These risky lifestyles, in turn, increase children’s likelihood of victimization (Osgood et al., 1996; Schreck et al., 2002; Tillyer et al., 2016; Turanovic et al., 2018). Utilizing social network analysis, Stogner et al. (2014) compellingly demonstrat that children’s own delinquent behavior such as heavy drinking and exposure to deviant peers significantly increase their odds of sexual victimization.
Finally, we include children’s exposure to general victimization as the third set of mediators. Several studies show that parental migration enhances children’s exposure to general victimization, such as bullying and physical victimization by strangers (Chen et al., 2020, 2017; Chen & Chan, 2016; Zhang et al., 2019). The temporal and geographical distance between parents and children left behind lowers the level of adult guardianship and increases children’s vulnerability to general victimization directly. Other environmental factors, such as increased familial stress and children’s involvement in risky lifestyles, can also elevate children’s level of vulnerability (Chen et al., 2017). Moreover, a large body of literature on repeat victimization indicates that exposure to general victimization increases children’s odds of other types of victimization, as the experience makes the victims more suitable targets in the eyes of potential offenders, or their ensuing behaviors—such as being aggressive and retaliatory—expose them to situations conducive to further victimization (M. Chen & Chan, 2016; Turanovic et al., 2018; Yan et al., 2018). Indeed, in a recent meta-analytic review, prior childhood sexual victimization is found to be the strongest predictor of children’s exposure to sexual assault (Assink et al., 2019).
Empirical Examination of the Association Between Parental Migration and Sexual Assault
As discussed above, the theoretical connections between parental migration and children’s sexual victimization are well established. Surprisingly though, there is little empirical evidence testing these theoretical propositions. The few studies that investigate this issue are either purely descriptive or having serious data limitations (e.g., using non-representative court cases). These studies, overall, provide anecdotal evidence that the prevalence of sexual victimization is relatively higher among LBC in rural China than other children (Jiang, 2012; Zhang, 2019; Zhang & Gen, 2016), and that the number of sexual victimization cases (e.g., rape and sexual harassment) against LBC has increased during the last 20 years (Jiang, 2012; Zhang & Gen, 2016).
Because of a lack of empirical data particularly large-scale, self-report data from the general population, researchers attempted to understand this critical social issue by analyzing publicly open court cases. These studies, in general, found a disproportionately higher number of sexual assault cases among LBC in rural China. Statistics from the court of the Zengdu district in Suizhou city, Hubei province, for instance, found that from the year 2006 to 2009, 95% of the victims of sexual assault against underage girls were rural LBC (Wang et al., 2020). Likewise, the local court of Sihong county, Jiangsu province, found that in the year of 2012, among 42 cases of raping and sexual harassment against women, 22 (52.3%) involved left-behind girls as victims (Zhang & Gen, 2016). Zhang (2019) used content analysis to analyze sexual abuse cases published online and found that migrant-sending provinces such as Guangxi and Hunan reported more of these cases. The official court data, however, severely underestimate rates of sexual victimization in rural China. The extensive practice of shaming and stigmatization against sexual abuse victims and their families particularly in rural China, coupled with complicated and long judicial procedures, prevent victims and their families from reporting these cases and seeking justice (Wang et al., 2020).
Data from the general population, however, contradict findings from the official court data. In the only two studies that directly investigated sexual victimization against LBC using adolescent samples (Chen & Chan, 2016; Yan et al., 2018), neither reported a significant association. Chen and Chan (2016) compared the prevalence of sexual victimization between children living with both parents and those left behind in rural China and found no significant difference between these two groups. Yan and colleagues (2018) collected childhood sexual victimization data among a sample of middle school students from Sichuan and Anhui provinces. Their results indicated that the prevalence of childhood sexual abuse was slightly higher (22.89%) among children left behind than children living with both parents (19.17%). This difference, however, was not statistically significant. These two studies, unfortunately, collected data from non-probability samples, making their findings difficult to be generalized to the general population.
Overall, prior research on the effects of parental migration on children’s sexual victimization in rural China is extremely limited. Few studies directly investigated this critical topic, and these studies, although informative, were hampered by severe data issues (e.g., non-representative samples) and produced largely inconsistent and sometimes conflicting findings. It is thus imperative to systematically investigate whether and how parental migration shapes the social constraints and opportunitities conducive to childhood sexual victimization in the unique cultural context of rural China.
The Context: Rural-to-urban Migration in Rural China
The urgency to study parental migration and its potential effects on LBC’s mental and behavioral outcomes is accentuated by China’s historically large-scale rural-to-urban migration and resulting social changes. Considered the largest migration in human history (Kong & Meng, 2010), 274 million people migrated from rural to urban areas in China in 2014 (National Bureau of Statistics of China [NBS], 2015). This number only slightly declined in the next few years and stood 244 million in 2018 (Department of Service and Management of Migrant Population National Health and Family Planning Commission of China, 2018). These migrants, mainly due to the restriction of China’s unique Household Registration (Hukou) system, often leave their children behind, creating a subpopulation of 61 million left-behind children (ACWF, 2013). It is estimated that 38% of children in rural China, and more than one-fifth of the child population overall, are LBC (ACWF, 2013), which leads to many anxieties and concerns that modern China is “raising a generation of left-behind children” (Stack 2010 ).
The unprecedented rural-to-urban migration in the last few decades has dramatically changed the family structure in rural China. Although the nuclear family is still the dominant form, its status has been steadily eroded since the 1980s, and the percentage of generation-skipping households—households with only grandparents and grandchildren—has dramatically increased (Hu & Peng, 2015). The rate of the generation-skipping household was estimated to be 3.26% in 2010, which was 3.37 times as many as in 1990 and 3.23 times as many as in 1982 (Hu & Peng, 2015). These statistics, however, may be conservative and overly underestimated. Many grandparents are primary caretakers of LBC but are not officially counted as members of the same household. National data, for instance, consistently show that a sizable proportion of children left behind are primarily cared for by grandparents (ACWF, 2013; Kong & Meng, 2010). A national survey revealed that among children left behind in rural China, 59% were cared for by grandparents, 35% were cared for by one parent left behind, and the remaining children were cared for by extended family members, friends, or lived by themselves (Kong & Meng, 2010). In other words, grandparent-led households (with both parents migrated) and single-parent-led households (with the other parent geographically separated) have increasingly replaced traditional nuclear families and become the normative forms of caretaking institutions in rural China.
In sum, such social changes at the macro-level as the massive migration and the resultant reconfiguration of family structure have substantially reshaped the lifestyles of and interactional patterns among children, parents, and extended family members (e.g., grandparents) at the micro-level, leading to potentially differential opportunities contributing to sexual victimization among child population in China. Against this backdrop, it is critical to systematically investigate a) whether China’s unprecedented rural-to-urban migration increases LBC’s exposure to sexual victimization, and b) whether the social mechanisms proposed by the Western victimization literature operate the same way in the unique context of rural China, a setting inherently different from Western societies and also undergoing dramatic and continuing structural and cultural transformations. Specifically, we hypothesize that
Hypothesis 1: Compared with children living with both parents, LBC have a higher likelihood of being sexually assaulted. Since we know little about whether different types of parental migration (i.e., maternal, paternal, and both-parent migration) influence children’s victimization dissimilarly, we do not specify potential differences across varying types of parental migration.
Hypothesis 2: Children’s bonding with primary institutions, involvement in risky lifestyles, and exposure to concurrent general victimization fully or partially mediate the association between parental migration and children’s risk of sexual assault.
We test these two hypotheses using a large-size, probability sample of middle school students in Guizhou province in southwest China.
Data and Methods
Data and Research Sites
The data for this study are from the Mental and Behavioral Health of Adolescents Survey, which collected information from a probability sample of middle school students in two districts of Guizhou Province in China: Guiyang and Qiannan Buyi and Miao Autonomous Prefecture. Conducted between November and December of 2019, the project aimed to examine the prevalence of children’s mental and behavioral problems (e.g., depression, delinquency, and exposure to victimization) and to identify individual and environmental factors associated with these varying outcomes.
We collected survey data from Guizhou province for two reasons. First, located in southwest China, Guizhou is one of the least developed regions in China and, thus, a primary migrant-sending province since China’s reform and opening-up in the 1980s, which is ideal for our research on the LBC population. Second, boasting 17 ethnic minorities, Guizhou province is one of the few multiethnic provinces from which a large probability sample of minority children can be readily drawn, who have often been overlooked by previous studies. We purposefully selected two districts, Guiyang and Qiannan Buyi and Miao Autonomous Prefecture, to capture the wide variations in socioeconomic status and ethnicity in this province. Guiyang district geographically encompasses Guiyang city, the capital of Guizhou province, and some boardering rural counties. It has a higher average income than the Qiannan Buyi and Miao Autonomous Prefecture and is dominated by the Han majority ethnicity. Qiannan Buyi and Miao Autonomous Prefecture, on the other hand, is remotely located in the southern part of Guizhou province and borders the Guangxi Zhuang Autonomous Region. It is less developed than Guiyang and inhabited mostly by minority groups such as Buyi and Miao ethnicity.
We adopted a multistage cluster sampling design to select our sample from the two districts. First, a list of all middle schools in Guiyang and Qiannan, which include seventh, eighth, and ninth grades, was obtained from each of the local Education Administrative Department. From this list, we randomly selected five middle schools in each city, with a total of 10 schools to participate in the study. In the second stage, we randomly selected two classes in each grade from the selected schools. These classes had a size of students ranging from 36 to 52. All students from the selected classes were invited to participate in the study. After obtaining consent from the teachers and adolescents themselves, we distributed questionnaires to participating students on regular school days, with teachers and school administrators not present on the site during the survey period. Researchers explained to the students the purpose of the research project and assured them that participation was voluntary, all personal information would be kept confidential, and students could withdraw from the study at any time.
In total, 2,533 students were selected to participate in the study. Thirteen students refused to participate, and 22 provided invalid data, which led to a total sample size of 2,498, registering a return rate of 98.6%. In terms of demographic characteristics, the average age of the sample was 13.5 years, with boys (n = 1,188) and girls (n = 1,189) equally split. The Han ethnicity accounted for less than half of the respondents in the sample (n = 1,083, 43.4%). Because this study focuses on children left behind in rural areas, we excluded urban students and rural-to-urban migrant students. The sample size of this study was reduced to 1,681. Table 1 provides a detailed description of each variable (mean, standard deviation, range, and number of valid cases) included in the data analysis.
Mean, Standard Deviation, Range, and Number of Valid Cases for All Variables.
Measurement
Dependent variable
The measure of our dependent variable sexual assault was derived from the Juvenile Victimization Survey (Finkelhor et al., 2005). The measure asked children in the last year whether a grown-up they knew or a stranger touched his/her private parts, made him/her touch their private parts, or forced him/her to have sex, whether a same-aged peer made him/her do sexual things, and whether they experienced an attempted or completed rape. The response categories for each of the four items were 1 = never, 2 = 1–2 times, 3 = 3–4 times, 4 = 5–7 times, and 5 = 8 times and more. The final scale of sexual assault, because of high skewness, was coded into a dummy variable, with 0 indicating that students answered “never” to each of the four items, and 1 otherwise. Previous research has demonstrated that this scale has good predictive validity and reliability in both western sample (Finkelhor et al., 2005) and Chinese sample (Chan, 2013).
Independent variables
The key independent variable in the study is parental migration status, which was assessed by asking children whether one of their parents currently worked in other cities or provinces or not. The variable was then recoded into four dummy variables, paternal migration, maternal migration, two-parent migration, and no migration. Sixty percent of the students reported that none of their parents migrated to other cities, 14% reported paternal migration, 7% maternal migration, and 19% dual-parent migration. It should also be noted that parental migration status often starts during children’s early childhood, and migrant parents visit home and their children infrequently due to time and financial constraints. Among students whose parents currently migrated, a predominant majority (93%) reported that their parents first migrated before they were 13 years old (the average age of our sample was 13.55), and the average age of parental migration was 6.64 years old. Also, about half reported that their parents had not visited them for half or less than half a year (53.6%), and 46.4% indicated that their parents had not visited them for more than half a year. These statistics suggest that for the predominant majority of the study students, their parent(s) migrated before the preceding year of the survey (i.e., the timeframe referenced for sexual assault).
Intervening variables
The study included a series of variables that hypothetically mediate the association between parental migration status and children’s exposure to sexual assault. First, we used caretaker monitoring and school bonding to capture children’s level of bonding to primary conventional institutions. Caretaker monitoring was measured by five items asking students how often their primary at-home caretakers knew where they were after school or when they went out at night, how often they told caretakers whom they were with when they were out, how often caretakers asked where they went, and whether they were expected to call when they were late (X. Chen et al., 2017). Response categories for each item ranged from 1 = never to 5 = often. The final scale of caretaker monitoring was computed by averaging the five items, with a higher score indicating stricter caretaker monitoring and supervision (α = 0.80). The second social control measure, school bonding, was measured by seven items asking students whether teachers in school cared about students, treated students fairly, were friendly to her/him, frequently communicated with him/her, and whether the student liked the campus environment, liked the school learning atmosphere, and studied hard (Shen & Zhong, 2018). Response categories for each item ranged from 1 = completely disagree to 5 = completely agree. Likewise, the final scale was computed by averaging the seven items, with a higher score indicating stronger bonding with schools (α = 0.86).
The second set of intervening variables tap into children’s delinquent and risky lifestyles, including measures of association with deviant peers, delinquent behavior, and unsupervised peer activities, which potentially elevate children’s exposure to opportunities conducive to sexual victimization. Association with deviant peers was a scale adapted from the National Youth Survey (Elliott & Huizinga, 1989), which consisted of seventeen items asking participants how many of their current friends participated in a series of delinquent activities in the last year (e.g., damaging school properties, stealing, carrying weapons to school, fighting, cheating on tests, group fighting, drinking, smoking, graffiti, truancy, and taking money from other people). For each item, response categories ranged from 1 = none of them to 5 = all of them. The average of these indicators was used to create a composite score for association with delinquent peers (α = .94). The second scale, delinquency, asked respondents how often they themselves were involved in each of these seventeen deviant or delinquent activities in the last year (Elliott & Ageton, 1980). The final scale was computed by averaging the seventeen items (α = .85). Finally, to capture children’s exposure to deviant opportunities, we created the scale unsupervised peer activities, which asked students how often they did things such as watching movies, riding bicycles, hanging around with friends, attending other gatherings or social activities, shopping and going out at night in the last month (Osgood et al., 1996). Response categories were 1 = never, 2 = 1–2 times, 3 = 3–4 times, 4 = 5–7 times, and 5 = 8 times and more. The final scale was computed by using the average of the six items (α = .78).
The third potential mediator was children’s experience of general victimization, measured by asking children whether they were robbed, assaulted with or without a weapon, or attacked but unsuccessfully, or whether their property was stolen or damaged intentionally by someone in the last year (Finkelhor et al., 2005). For each item, response categories were 1 = never, 2 = 1–2 times, 3 = 3–4 times, 4 = 5–7 times, and 5 = 8 times and more. The final scale was computed by using the average of the six items (α = .79).
Control variables
A series of individual and household characteristics associated with parental migration and risk of sexual assault were controlled to account for selection bias. Specifically, we controlled for children’s individual characteristics such as age, gender (0 = female, 1 = male), ethnicity (0 = minority ethnicity, 1 = Han), and physical health (ranging from 1 = very bad to 5 = very healthy), as well as family characteristics including parental education (measured as the average of mother’s and father’s education), children’s perceived family economic status (ranging from 1 = not good to 5 = very good), whether biological parents are divorced or not (0 = no, 1 = yes), and the number of siblings. The descriptive statistics (mean, standard deviation, and range) for each of the variables in the analysis are presented in Table 1.
Analytic Strategy
We began our analysis by first investigating the bivariate association between parental migration status and children’s exposure to sexual assault. After the bivariate association was established, we ran a series of logistic regression models to explore whether the association remained robust when individual and familial characteristics were controlled, and whether children’s bonding with conventional institutions, risky lifestyles, and experience of general victimization mediated this association. The direct comparison between uncontrolled and controlled regression coefficients to assess potential mediating effects, a common practice in nested linear regression models, however, is not legitimate in non-linear logit and probit models (Karlson et al., 2012). Following the suggestion by Karlson and colleagues (Karlson et al., 2012), we remedied this problem by running additional analyses in which original mediators are replaced by x-residualized mediators (rescaled models in Table 3). To assess the effects of potential mediators, we then compared regression coefficients of parental migration status between models with original mediators and their counterparts with x-residualized mediators.
To formally test whether the mediating effects are statistically significant or not, we adopted the bootstrapping approach suggested by Hayes (2017) to calculate indirect effects for each type of parental migration status on sexual assault. Finally, we applied the Full Information Maximum Likelihood (FIML) approach to address missing data, which is proved to provide efficient estimations of statistical parameters and less biased estimates of standard errors compared with other missing data imputation methods (Schafer, 1997). All our multivariate analyses were conducted in the statistical program Mplus7.4 (Muthén & Muthén, 2012).
Results
Bivariate Analysis
Table 2 illustrates the bivariate association between parental migration status and children’s exposure to sexual assault. Overall, 7% of children in rural China reported that they experienced at least one type of sexual assault in the last year. More specifically, 3.1% of these children reported that their private parts were touched by a known adult, or they were asked to touch the adult’s private parts or were forced to have sex with the known adult. Close to three percent (2.9%) reported that they had the same experience with an unknown adult. Moreover, 3.5% of the children reported that they were made to do sexual things with peers, and 2.7% reported being raped or attempted rape. In terms of the relationship between parental migration status and sexual assault, the results show that children with a migrant mother (maternal migration) reported the highest rates for the overall scale of sexual assault and for each individual item constituting the scale. Children with both parents migrating reported the second-highest rates. The difference between paternal migration and non-migration, however, was small and inconsistent across items. The associations between parental migration status and children’s exposure to sexual assault, measured as an overall scale or measured using each individual item, were all statistically significant at 0.05 level.
Bivariate Association Between Parental Migration Status and Children’s Exposure to Sexual Assault (%).
Note. n = 1681.
Logistic Regression Analysis
We investigated the effects of parental migration on children’s exposure to sexual assault and the mediating roles of bonding with conventional institutions, risky lifestyles, and prior general victimization in a series of logistic regression models (Table 3). Model one demonstrated that parental migration status was significantly associated with children’s experience of sexual assault when individual and family background were controlled. Compared with the reference group (none of the parents migrated), maternal migration (exp(b) = 2.30, p < .01) and both-parent migration (exp(b) = 1.96, p < .01) increased children’s odds of sexual assault significantly. Paternal migration was also associated with an increased probability of sexual assault, this association, however, was not statistically significant (exp(b) = 1.25, p > .05). Among the control variables, children’s health was a significant predictor, with better health predicting decreased odds of sexual assault (exp(b) = 0.15, p < .05).
Logistic Regression Models Predicting Sexual Assault Among Children in Rural China.
Note. n = 1681. *p < .05. **p < .01.
We included children’s bonding with school and caretaker monitoring in Model two to assess their potential mediating effects. Results indicated that both predictors were statistically significant, with the odds of being sexually assaulted decreasing by 27% for each unit increase of school bonding (exp(b) = 0.73, p < .01) and by 30% for each unit increase of caretaker monitoring (exp(b) = .70, p < .01). The association between parental migration status and sexual assault also appeared to be partially mediated. The absolute value of the regression coefficient for maternal migration decreased from 0.92 (Model 2 rescaled) to 0.70 (Model 2), a 24% decrease, and the regression coefficient for two-parent migration decreased from 0.72 (Model 2 rescaled) to 0.53 (Model 2), a 26% decrease.
Children’s risky lifestyle measures such as association with delinquent peers, unsupervised peer activities, and engagement in delinquency were stepped into Model three. All the three measures were statistically significant. A one unit increase in peer delinquency increased children’s odds of sexual victimization by 1.89 times (exp(b) = 1.89, p < .01). Children’s own delinquent behavior and time spent on unsupervised peer activities had a similar effect on sexual victimization (exp(b) = 1.64, p < .01 and exp(b) = 1.59, p < .01, respectively). The comparison between models with original mediators and x-residualized mediators suggests that the lifestyle measures had a sizeable mediating effect. The absolute value of the regression coefficient for maternal migration decreased from 1.05 (Model 3 rescaled) to 0.73 (Model 3), a 30% decrease, and the regression coefficient for two-parent migration decreased from 0.81 (Model 2 rescaled) to 0.50 (Model 2), a 38% decrease. The effect of two-parent migration also decreased to non-significance in model 3.
In the final model, we added children’s experience of general victimization (Model 4). The results showed that the effect of maternal migration, after controlling for all the individual and environmental factors, remained statistically significant. The coefficient for both-parent migration was non-significant and decreased substantially from 0.79 (Model 4 rescaled) to 0.43 (Model 4), a 46% decrease. Children’s exposure to general victimization was a significant predictor associated with increased odds of sexual assault (exp(b) = 1.63, p < .01). Moreover, three other intermediate environmental factors remained to be statistically significant. Caretaker monitoring served as a protective factor against sexual victimization (exp(b) = 0.72, p < .01), and peer delinquency (exp(b) = 1.68, p < .01) and time spent on unsupervised activities (exp(b) = 1.52, p < .01) elevated children’s odds of being sexually assaulted.
Mediation Analysis
Whereas our regression models suggest that a series of environmental factors collectively mediate the association between parental migration and children’s exposure to sexual assault, it is not clear which individual risk and protective factor serves as a mediator and how strong the mediating effect is. To formally test the roles of these intervening variables, we used the bias-corrected bootstrapping method with 5,000 resampling (Hayes, 2017) to calculate parameter estimations and confidence intervals for the total, direct, and indirect effects for each type of parental migration status (Table 4). More specifically, we used measures of children’s bonding with primary conventional institutions, risky lifestyles, and exposure to general victimization as parallel mediators, parental migration status as the predictor, and sexual assault as the outcome variable. The same set of control variables in regression models were used to predict all the mediators and the outcome variable (Hayes, 2017).
Total Effects and Indirect Effects of Parental Migration on Sexual Assault Using Bootstrapping Method.
Note. n = 1681.
The bold entries indicate that the confidence interval is statistically signficant from zero.
The mediation effect analyses provide additional insight into the hypothesized association between parental migration and children’s exposure to sexual assault. Consistent with our regression models, the mediation analysis revealed that paternal migration did not have a significant total effect on children’s sexual victimization. Paternal migration, however, indirectly increased children’s probability of sexual victimization through two pathways: weakened caretaker monitoring and supervision (b = 0.03, 95% C.I [(0.01, 0.06]), and increased exposure to general victimization (b = 0.01, 95% C.I [0.00, 0.03]). Moreover, the results indicated that maternal migration had an overall significant effect on children’s victimization (b = 0.52, 95% C.I [0.15, 0.82]), with both direct and indirect effects statistically significant. The indirect effect went through three primary pathways: weakened caretaker monitoring (b = 0.06, 95% C.I [0.02, 0.11]), peer delinquency (b = 0.06, 95% C.I ([0.02, 0.12]), and general victimization (b = 0.04, 95% C.I [0.01, 0.08]). Finally, the total effect of both-parent migration on sexual assault was statistically significant, albeit slightly smaller than that of the maternal migration (b = 0.40, 95% C.I [0.17, 0.63]). The mediating social processes and their effect sizes for both-parent migration were almost identical to those of maternal migration. That is, the difference in sexual victimization between both-parent migration and non-migration household was due to weakened caretaker monitoring (b = 0.05, 95% C.I [0.02, 0.09]), increased association with delinquent peers (b = 0.07, 95% C.I [0.03, 0.11]), and elevated exposure to general victimization (b = 0.03, 95% C.I [0.01, 0.06]).
Sensitivity Analysis
In light of the considerable diversity in the sample and the importance of gender and race/ethnicity in the sexual victimization literature (Assink et al., 2019), we further conducted sensitivity analyses to examine whether our findings are robust across ethnic and gender groups. Our gender-specific analyses revealed that age was the only variable that differentially predicted sexual assault, with older girls more likely to be sexually assaulted (b = 0.29, p < .10) and older boys less likely to be victimized (b = –0.19, p > .05). In terms of ethnicity, exposure to general victimization significantly increased odds of sexual assault among children of Han ethnicity (b = 1.06, p < .01), but had no effect among minority children. The equality of regression coefficient test (Brame et al., 1998) revealed that the effects of general victimization differed significantly across the two groups (z = –2.72, p < .01). Overall, whereas the significance levels of some predictors slightly differed across gender and ethnicity subgroups, substantive findings held essentially the same, suggesting that our results derived from the full sample were robust across gender and ethnic subgroups. These findings were uploaded as an appendix table to save space (Online Appendix Table 1).
Additionally, we performed sensitivity analyses for each of the individual items of sexual assault (i.e., sexual assault by known adults, by adult strangers, by peers, and rape/attempted rape) by re-running the baseline and the full models in Table 3 (results are in Online Appendix Table 2). The nuanced analyses provide substantively similar findings as those reported in Table 3. Specifically, maternal migration and both-parent migration elevated children’s odds of being sexually assaulted, and these relationships can be largely explained by a lack of caretaker monitoring, children’s increased association with delinquent peers, higher levels of involvement in delinquency, more unsupervised leisure time, and hightened exposure to general victimization.
Discussion and Conclusion
This study furthers understanding of sexual assault against children left behind in rural China by quantitatively assessing its prevalence and identifying underlying social processes. This focus is particularly compelling because China’s unprecedented rural-to-urban migration has created a large population of LBC yet sexual victimization in this highly disadvantaged group eludes empirical research. Whereas a small number of descriptive studies provide anecdotal evidence that LBC have a high risk of sexual victimization, findings from these studies are largely inconclusive. Using a representative sample of rural middle school students, this study demonstrates that parental migration, overall, increases children’s exposure to sexual victimization. Moreover, the associations between parental migration and children’s exposure to sexual assault are partially or fully mediated by children’s bonding with primary institutions, their risky lifestyles, and their exposure to general victimization.
Consistent with our first hypothesis, our results reveal that parental migration is associated with LBC’s increased odds of sexual assault in rural China. Specifically, maternal migration and both-parent migration significantly increase LBC’s vulnerability to sexual victimization. Our results contradict the null findings in two prior studies that similarly examined parental migration and sexual victimization (Chen & Chan, 2016; Yan et al., 2018) but are consistent with the larger literature, which suggests that parental migration is associated with a myriad of poor mental and behavioral outcomes among children left behind (Parreñas, 2005b; Qin & Albin, 2010; Sun et al., 2015; Toyota et al., 2007). The absence of parents as key guardians, as well as the cumulative effects of parental migration on children’s interaction with external environment, provides predators increased opportunities to exploit this vulnerable population.
Adding to a small but growing literature (Chen & Jiang, 2019; Liu et al., 2009; Parrenas, 2009; Wen & Lin, 2012; Xu et al., 2019), our study provides strong evidence for the differential effects of parental migration and caretaking arrangements on children’s likelihood of sexual assault. In traditional societies such as rural China, the deeply embedded social norms regarding the division of labor in a family—with the mother as a care provider and the father as a breadwinner—shape and structure the interactional patterns among migrant parents, left-behind caretakers, and children (Abbott et al., 1992; Parreñas, 2005b). When these social norms are not violated, as in the case of paternal migration, children’s normative mental and physical development is less likely to be disrupted. In contrast, maternal migration (and both-parent migration) openly contests this deeply rooted social norms, thus imposing significantly deleterious effects on children’s risk of sexual assault (Liu et al., 2009; Oliveira, 2018; Parreñas, 2005b). Echoing early studies (Miethe & McDowall, 1993; Sampson & Lauritsen, 1994), these findings demonstrate that the intersections among social changes, lagged social norms, and individual and familial adaptations shape and structure children’s differential vulnerabilities to sexual victimization in rural China.
Our second hypothesis regarding the mediating effects of children’s bonding with conventional institutions, engagement in risky lifestyles, and exposure to general victimization is also empirically supported. Supporting the transitional perspective, our findings reveal that whereas paternal migration does not have a significant total effect, it does indirectly enhance children’s likelihood of sexual victimization. There are two primary pathways: a lack of parental monitoring and discipline, and children’s increased exposure to general victimization. In other words, even when a mother is left behind to serve as the primary caretaker—which conforms to the social norms regarding the gendered division of labor—parental migration still creates unique opportunities and pathways through which the left-behind children’s odds of sexual victimization can be amplified. This indirect effect, however, may be countervailed by other social processes, making the total effect of paternal migration small and non-significant.
Our results also demonstrate that the effect of maternal migration is partially mediated by three intervening factors: reduced caretaker monitoring and supervision, increased association with deviant peers, and elevated exposure to general victimization. Previous studies point out that whilst a migrant mother is physically distant, the expectations for her to fulfill maternal duties from family members, particularly from the husband, are not fundamentally modified (Oliveira, 2018; Parreñas, 2005b). As a result, the left-behind father often struggles to adapt to the reconfiguration of the family structure and to embrace dual-parent duties, leading to a lack of supervision and monitoring over children and tense relationship between fathers and children (Parreñas, 2005b). Children left behind, at the same time, may resent being “abandoned” by their mothers and the resultant “minimized mothering” (Oliveira, 2018; Parreñas, 2005b), and respond by increasingly associating with deviant friends for companionship and emotional support, elevating their odds of victimization in general and sexual assault in particular (Chen et al., 2017).
Moreover, the direct association between maternal migration and children’s sexual assault remains statistically significant when the proposed intervening social factors are taken into account, suggesting that other social processes are simultaneously at work. Previous studies reveal that sexual assault in rural China is often committed by family members or other adults in the same village (Jiang, 2012; Zhang, 2019; Zhang & Gen, 2016). Maternal migration, thus, may exaggerate the prominence of this social mechanism, leaving children left behind more vulnerable to assault of intimate family members or acquaintances in the absence of a mother’s guardianship and protection.
Our finding that children from both-parent migrant households have a higher likelihood to be sexually assaulted than those from non-migrant households is also expected. Like maternal migration, three primary pathways, including reduced caretaker monitoring and supervision, more frequent association with deviant friends, and increased exposure to general victimization, operate collectively to promote a higher risk of sexual victimization against children from both-parent migrant households. Considering that grandparents are most frequently the primary caretakers for this group (Chen et al., 2011), our results imply that grandparents—whose role is to provide emotional affection instead of discipline and punishment pre-migration—are ill-equipped to serve as surrogate parents and to shield children from sexual victimization. Indeed, many grandparents report that they are frustrated and feel helpless when grandchildren intentionally disobey or misbehave (Zhang, 2011). The lack of monitoring and supervision, as well as children’s involvement in risky lifestyles, in turn, increase children’s risk of general victimization and sexual assault. This finding, however, is inconsistent with a recent study (Chen et al., 2017), which reports that children living with grandparents have a similar level of general victimization compared with those living with non-migrant parents. Although it is possible that risk factors contributing to sexual victimization are uniquely different from other types of victimization, more future studies are needed to confirm whether our results can be replicated and generalized to a larger population.
It may also be difficult to grasp the finding that the effect size of both-parent migration on sexual assault is slightly smaller than that of maternal migration. We speculate that the income differential between one- versus both-parent migration and, consequentially, the differential capacity of the family to formally or informally compensate caretakers’ financial loss plays a critical role. A rural household’s financial gain, as a unit, is maximized when both parents migrate and grandparents and children stay behind. This arrangement enables grandparents to regularly receive remittance from adult migrant children, thus significantly cutting off their farming-related duties and facilitating their commitment to full-time caretaking (Asis, 2006; Ye & Murray, 2005). In contrast, in the scenario of maternal migration, the left-behind father is the sole primary caretaker, who often struggles to perform dual-parent duties and only receives assistance from grandparents sporadically. In other words, compared with grandparent caretakers, a stay-behind father often shoulders a significantly heavier load of household responsibilities (Chang et al., 2011), often along with his regular work responsibilities, and spends less time with their children, which elevates children’s risk of victimization. Recent studies, however, point out that there is much heterogeneity in the grandparent caretaker group (Chen, under review), thus it is important to investigate how different arrangements within this marginalized group impact the likelihood of children’s sexual victimization.
In summary, these results highlight the detrimental effects of parental migration on childhood sexual victimization in rural China. Some limitations should be addressed before discussing the policy implications of our study. First, our study focuses on the protective and risk factors of sexual victimization at the individual level against the larger context of urbanization and migration at the macro level. Due to a lack of data, it does not take into account exosystem level factors such as community or neighborhood characteristics (e.g., percentage of parental migration in a village, area crime rate). Future studies, upon the availability of data, need to explore whether factors at different levels cross-interact with each other to create social constraints against or opportunities conducive to childhood sexual victimization. Second, our study uses parental migration as a proximate indicator of caretaking arrangement, which omits the possibility that there may exist different caretaking arrangements for the same type of parental migration. For example, in both-parent migrant households, children may be primarily cared for by grandparents or by other extended family members. The latter group of children potentially faces a higher risk of victimization, considering that caretaking by family members other than parents and grandparents is not culturally promoted or endorsed and children under this type of care may be double disadvantaged. Our inclusion of the caretaker-child bonding and monitoring measures, which are more proximal predictors of delinquency and victimization than caretaker arrangement, should substantially mitigate this measurement limitation. Nonetheless, future studies need to create more detailed classification of caretaking arrangements, ideally with a large sample size, perhaps with the grandparents-as-caregiver group oversampled. Survey data should also be triangulated with qualitative data such as interviews and case studies to enrich the findings. Third, the study uses cross-sectional data and cannot establish a temporal order for some of the relationships assessed. For example, it is possible that experience of sexual assault can increase juveniles’ engagement in delinquent behaviors and association with delinquent peers (e.g., Hay & Evans, 2006). We are fairly positive about the temporal order between the independent and outcome variables of parental migration and children’s sexual victimization though, as the victimization measure has a time frame set to “during the past year” yet parental migration status often starts earlier than the preceding year of the survey. In fact, in migrant-sending areas such as Guizhou province, rural-to urban migration has been a rite of passage for rural young people, and people often migrate before they are married or during their children’s early childhood (Ye, 2018), including those in our sample. As previously mentioned, for a predominant majority of our study sample, parental migration precedes children’s victimization in time. That being said, longitudinal studies are still needed to establish more definite temporal orders among the study variables and to examine the longitudinal association between parental migration and left-behind children’s risk of sexual victimization.
Despite limitations, results of this study carry important policy implications. Adding to the prior literature, our findings highlight the importance of paying heed to the social costs of China’s massive rural-to-urban migration. When China’s urban middle-class and elite class residents take advantage of the so-called “population bonus,” namely the low cost of rural-to-urban migrant labor, the fact that migrant parents and their children disproportionally shoulder these social costs is deliberately overlooked. The sacrifices endured by migrant parents and their children need to be immediately recognized and addressed by society. With the irreversible trend of urbanization in China, it is imperative that the central government implements policies to eliminate the enduring parent-children separation in rural China by dramatically reforming the outdated Hukou policy and providing migrants and their children access to public resources such as subsidized housing, public schools, and urban medical care in migrant-host cities. At the local level, the state and local education system needs to develop educational curriculum in primary and middle schools to address sexual victimization, elevating children and caretakers’ awareness of sexual victimization risk and implementing evidence-based strategies to effectively protect children from potential offenders and to cope with traumatic victimization experiences. Finally, the findings of this study can be utilized to improve rural parents’ understanding of the undesirable effects of parental migration, particularly maternal migration, and enable them to make a more informed decision about migration and postmigration caretaking arrangements.
In conclusion, our study contributes to the literature by being the first one to systematically investigate the prevalence and etiology of sexual assault against children left behind in rural China. Overall, our findings confirm that parental migration, particularly maternal and both-parent migration, significantly increase children’s risk of sexual assault. The effects of parental migration are mediated by three primary pathways: weakened caretaker monitoring and supervision, children’s frequent engagement in risky lifestyles, and their elevated exposure to general victimization. These findings highlight the social costs of China’s dramatic rural-to-urban migration in the last three decades and the urgency to develop prevention and treatment programs based on a holistic understanding of key structural, cultural, and individual protective and risk factors of sexual assault against LBC.
Supplemental Material
sj-docx-1-cad-10.1177_0011128721989072 – Supplemental material for Parental Migration and Risk of Sexual Assault Against Children in Rural China
Supplemental material, sj-docx-1-cad-10.1177_0011128721989072 for Parental Migration and Risk of Sexual Assault Against Children in Rural China by Xiaojin Chen, Yuning Wu and Jia Qu in Crime & Delinquency
Footnotes
Authors’ Note
This article was based on data collected by the project “Studying Poly Victimization against Children Left Behind in Rural Guizhou and Its Prevention.”
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The project was supported by a theoretical innovation grant awarded by the Guizhou Social Science Association (GZLCLH-2019-011).
Supplemental Material
Supplemental material for this article is available online.
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
