Parents Returning to Parents: Does Migration Background Have an Influence on the “Boomerang Effect” Among Parents After Divorce?

The literature on the “boomerang effect” (returning to the parental home) is predominantly focused on young adults returning to their parents’ home in response to economic hardship. In this chapter, we study the boomerang effect in the context of divorce by looking at who returns to the parental home after a relationship break-up. The analysis is restricted to individuals with children, as the presence of children in the dissolving household gives an additional dimension to the process of young adults moving back in with their parents. We use register data from Belgium to test the probability of returning to the parental home among individuals who were either married or cohabiting. In addition, we take population heterogeneity into account as we compare the boomerang effect among different origin groups (Belgians, Turks, Moroccans). We find that the boomerang process is highly gendered, with men being substantially more likely than women to return to the parental home. Our results also indicate that migration background plays a less important role, with the exception that fathers of Turkish and Moroccan origin are more likely than others to return to the parental home after separation.


Introduction
Life course studies have often investigated the questions of why, and, if so, to what extent adult children return to the parental home. The economic recession of 2008 has sparked a renewed interest in understanding how socio-economic conditions and welfare state policies influence this so-called "boomerang effect" (Arundel and Lennartz 2017). In addition to job loss, divorce or relationship dissolution is generally considered to be one of the major life events that can cause adults to return to their parents' home. Surprisingly, most existing studies on this topic have focused on young adults. Thus, the boomerang effect is placed in a context of young adults returning soon after leaving the parental home to gain independence from their parents or in response parent-child conflicts. Only a few studies (e.g., Albertini et al. 2018) have examined the boomerang effect among adults who are somewhat older, and who use returning to the parental home as a post-divorce housing strategy. Moreover, even fewer studies have taken parenthood among divorcees into account when investigating the return to the parental home after a break-up (Guzzo 2016).
A second gap in the existing literature on the boomerang effect is that most studies do not consider population heterogeneity in migration status and origin group. It is, however, important to take this heterogeneity into account given the increasing diversity in most societies, and the differences between migrant and native-born populations in terms of family patterns, socio-economic position, and family attitudes. Research has also shown that acceptance of union dissolution -and of divorce in particular -is much lower in certain minority groups than it is the majority population (Koelet et al. 2009a). In addition, there is evidence that in Europe, some migrant groups have more universalistic family values than native populations, who tend to display more individualistic value patterns (Merz et al. 2009;Phalet and Schönpflug 2001). It has, for example, been shown that if an individual has to leave the marital home due to divorce, his/her ethnic background will greatly affect his/ her likelihood of returning to the parental home (Kleinepier et al. 2017). Whether this effect differs depending on the individual's characteristics prior to the union dissolution and the characteristics of the parental household remains to be investigated.
While taking migrant populations into account, this chapter poses two main research questions. First, we investigate whether having a migrant background influences the likelihood among mothers and fathers of returning to the parental home after a relationship break-up (marriage or cohabitation). We look at how the boomerang effect plays out differently in the lives of fathers and mothers of Belgian, Turkish, and Moroccan origin, and take the migration background of the ex-partner into account. Second, we investigate whether economic and family differentials within these origin groups influence the likelihood of returning to the parental home (Mortelmans et al. 2019).
To answer these research questions, we use data from the Crossroads Bank of Social Security. These register data contain three "dissolving cohorts" (separation in 2007, 2008, or 2009). In the following, we provide an overview of the literature on the boomerang effect in the context of divorce and separation, and then present hypotheses that address our central research questions.

Prior Research and Hypotheses
In general, men are more likely than women to return to the parental home after a divorce or the end of a cohabiting partnership. Among the potential explanations for this gender difference are that the presence of children increases the likelihood that women will stay in the previously shared home after the relationship ends (Ongaro et al. 2008;Stone et al. 2014;Sullivan 2007). It has, conversely, also been argued that women are more likely than men to turn to their parents for help, and that daughters are more likely than sons to receive assistance from their parents, especially in times of need (Guzzo 2016). Additionally, there is evidence that adult children with children of their own receive more support from their parents than childless adults (Fingerman et al. 2009). As women are more likely than men to be living with their children after leaving a relationship, it may be expected that women would have higher odds of returning to the parental home (Guzzo 2016). However, most empirical studies have provided support for the hypothesis that men have a greater tendency than women to return home. Stone et al. (2014) found that in the UK, mothers were significantly less likely than fathers to return to the parental home after relationship dissolution. Meanwhile, Guzzo (2016) found evidence that led them to reject their hypothesis that mothers have an increased likelihood of moving back in with their parents. Based on these empirical findings, we formulate the following hypothesis: Hypothesis 1: Net of socio-demographic and economic characteristics, fathers are more likely than mothers to return to their parents' home after relationship dissolution, regardless of whether they have a migrant background.
Although research on the consequences of divorce among couples with a migrant background is limited, the existing studies on this topic have shown that the likelihood of returning to the parental home after divorce may differ by migrant background. Whereas adult children having autonomy and being independent of their parents are norms in Western societies (Billari et al. 2001), these norms appear to be less prevalent among minority groups with a non-Western background (Merz et al. 2009;Phalet and Schönpflug 2001). Family relationships, community ties, and intergenerational support are often emphasised among these groups. In addition, both first-and second-generation migrants tend to lean on support from informal networks in all domains of life, such as in finding a job, organising informal childcare, and finding a suitable partner. Receiving informal support is particularly important for subpopulations who have limited educational and labour market opportunities, such as individuals with a Turkish or a Moroccan background living in Belgium. Men and women who are first-or second-generation Turkish or Moroccan migrants often find themselves in a precarious socio-economic position because they tend to have limited opportunities, while also facing relatively high levels of discrimination on the labour and housing markets (Baert et al. 2015;Zick et al. 2008). A qualitative study by Koelet et al. (2009a) conducted among divorced men and women who were second-generation Turkish or Moroccan migrants showed that even though their parents often disapproved of divorce as a solution for relationship problems, their parents continued to support them. The results also showed that divorcees of Turkish or Moroccan origin often returned to the parental home as a strategy for coping with financial difficulties. Moreover, the study found that receiving informal support, mainly from the divorcees' immediate family, was often a stepping stone to receiving formal support. An analysis of quantitative data from the Belgian National Register showed that the odds of returning to the parental home were highest among Turkish men (Koelet et al. 2009a). Our hypothesis is, therefore, as follows: Hypothesis 2: The probability of returning to the parental home after divorce or separation is higher for Turks and Moroccans than it is for Belgians.
In addition, the likelihood of returning home might differ not just between men and women or between people with different migrant backgrounds, but between men and women within these communities. Koelet et al. (2009b) found that although the Turkish and the Moroccan communities are generally more disapproving of divorce than the Belgian community, the stigma is more severe for women than for men. As maintaining marital satisfaction is often seen as the wife's responsibility, women are more likely than men to be blamed by the community for the failure of a marriage. Hence, parent-child conflict is expected to arise more frequently for women than for men with a Turkish or a Moroccan background. Based on these assumptions, our hypothesis is a follows: Hypothesis 3: Mothers with a Moroccan or a Turkish background are less likely to return home after divorce or separation than fathers with the same background.
In addition to differences by gender, there are several other explanations for why some types of parents are more likely than others to move back in with their parents after a divorce. Access to financial resources is an important factor in whether adult children return to the parental home (Goldscheider and Goldscheider 1998;Whittington and Peters 1996). Paying not just for housing, but for utilities, food, and other basics of daily living can be difficult for single individuals (Furstenberg et al. 2005). Thus, economic necessity may drive these newly single people back into their parents' home. This observation leads us to formulate the following hypothesis: Hypothesis 4: Individual income is negatively associated with the probability of returning to the parental home after divorce or separation, regardless of the person's gender or migrant background.
Unemployment and inactivity restrict people's residential independence in two ways. First, if individuals lack economic resources because they have no labour market income, they may be unable to pay rent or make mortgage payments. Moreover, people who lack steady employment and adequate income face barriers to being accepted as a tenant (Loopmans et al. 2014) or to obtaining a loan, especially since the economic recession of 2008. However, net of the income effects that are inherent in an individual's employment status, the issue of childcare arises for parents who are employed. Men and women with children who are employed fulltime may benefit from the childcare that their own parents (i.e., their children's grandparents) can provide. Guzzo (2016) and Kleinepier et al. (2017) found that parents with year-round, full-time employment were the least likely to return home. These studies did not, however, specifically identify parents with different employment patterns, and they did not account for personal income. We therefore propose the following hypothesis: Hypothesis 5: Net of income effects, employment is positively associated with the probability of parents returning to the parental home after divorce or separation, regardless of their gender or migrant background.
Although we predict that, in general, returning to the parental home has a negative association with income, there is one low-income group who might be less prone to return to the parental home: parents who receive welfare benefits. At the macro level, Arundel and Lennartz (2017) argued that the likelihood of returning home is lower in more protective welfare states such as Belgium. At the micro level, there are several explanations for why welfare dependency is negatively related to returning to the parental home. First, recipients of means-tested benefits might find that they would lose the benefits they are currently receiving if they moved in with their parents, as the combined household income would render them ineligible. Second, individuals who receive welfare benefits have better access to affordable, social, and state-provided housing than non-recipients. In addition, single parents receive more benefits than childless singles, and they are often given priority in the allocation of social housing. As women are more likely than men to be the primary carers for their children, we test the following hypotheses: Hypothesis 6: Net of income and employment effects, welfare dependency is (a) associated with a lower likelihood of returning to the parental home after divorce or separation. This association is (b) weaker for men than it is for women.
In addition to generally assuming that returning to the parental home is more common among second-generation Moroccan and Turkish migrants than among the native population, we expect to find variation within each origin group. First, we expect to observe differences depending on the relationship type of the ex-couples. Among Moroccan and Turkish origin groups, the majority of unions are marriages, and unmarried cohabitation is less common. Qualitative research by Koelet et al. (2009b) has shown that the parents of second-generation Turkish and Moroccan migrants are less supportive of unmarried cohabitation, and generally expect a serious relationship to result in engagement and marriage. Additionally, secondgeneration migrants often adhere to the traditional views on union formation that are dominant in their parents' countries of origin (De Valk and Liefbroer 2007). When a man or -in particular -a woman from a migrant community opts for unmarried cohabitation instead of marriage, it is often a strong indication of the individual's independence from the influence of his/her parents, given that this choice conflicts with their community's norms on union formation. In light of the marriage norm among these migrant groups and the implications of a failure to adhere to it, we test the following hypothesis: Hypothesis 7: The probability of returning to the parental home after divorce or separation is higher for Turks and Moroccans who were formerly married rather than cohabiting than it is for Belgians.
Second, we expect to find differences in the propensity to return to the parental home depending on the migrant status of the partner. While the number of marriage migrations has decreased considerably over the past decade (Dupont et al. 2017), substantial shares of second-generation Turks and Moroccans have married a firstgeneration partner from their country of origin (Huschek et al. 2012). For secondgeneration parents of Turkish and Moroccan origin, choosing a first-generation partner from their country and/or region of origin is often preferred to choosing a partner from the second generation or from another origin group (Koelet et al. 2009a). In addition, an individual who chooses a partner from his/her country of origin generally has strong connections with his/her migrant community in both in the country of residence and the country of origin (Lievens 1997), which can, in turn, strengthen the individual's bonds with his/her family and the broader migrant community. The decision not to follow the partner choice expectations of one's parents and community can be an indication that an individual has a greater need for independence or a greater cultural distance from his/her home country. We therefore propose the following hypothesis:

Hypothesis 8: The probability of returning to the parental home is higher for Turkish
and Moroccan ex-partners whose former spouse was a first-generation migrant.

Data
In this study, we used Belgian data from the Data Warehouse on Labour Market and Social Security. This large-scale administrative dataset contains information from nearly all social security agencies in Belgium (e.g., the National Office of Social Security, the National Employment Office, and the National Institute for Health and Disability Insurance). A sample was drawn that consists of 46,050 households who had experienced the dissolution of their relationship in 2007, 2008, or 2009. In the registers, the definition of a household was based on co-residence. The sample unit was based on the relationship status of couples in a household at moment t (2007-2008-2009) compared to their status at t−1. The use of this approach allowed us to draw a sample of 21,600 divorced couples (at time t) and 24,450 separated (and formerly cohabitating) couples (at time t). To enable us to study differentials by origin group, the sample was stratified by the (non)-migrant status (Belgian, European, or non-European) and gender of at least one of the (ex-)partners. "Migrant status" is defined as being a first-, second-(parents), or third-(grandparents) generation migrant based on the country of birth of the individual and of his/her (grand) parents. If no information on the country of birth was available, we used an individual's first nationality as an alternative indicator of his/her origin. For the Belgian sample members, we added the requirement that the respondent's partner also had a non-migrant Belgian background. For couples in which one of the partners had a migrant background, no restrictions regarding the migrant background of the other partner were applied. This sampling strategy resulted in 30,000 couples with two Belgian partners, 3000 couples with a European woman, 6500 couples with a European man, 3000 couples with a non-European woman, and 3550 couples with a non-European man. As each household had to be married or cohabiting at t-1, the total number of individuals in the sample was 92,100 (each of the 46,050 sample members and his/her respective (ex-)partner). Recent migration movements are not immediately registered in the Data Warehouse. Since the most recent data in the study were from 2013 (t + 4 in the 2009-wave), we assume that migrants from the most recent wave are not included in our sample. Moreover, as illegal migration is not covered by the administrative data, undocumented migrants could not be included in the models. The data provide yearly information on both the ex-partners and their household members after union dissolution. For this chapter, we limited the sample to mothers and fathers who were of Belgian origin or Turkish or Moroccan migrants of the second generation or later. The respondents' parental status was determined based on the presence of at least one dependent child in the household at t−1. We excluded ex-partners who were first-generation immigrants. Since first-generation migrants often come to Belgium without their parents, including them would have posed a measurement problem in our models. As we have no information on a potential return to the parental home in the country of origin, we would have underestimated the return parameter. For migrants of the second or a subsequent generation, we could be certain that their parents were present in Belgium, and that data on them would be available in the registers. In addition, self-employed individuals were excluded from the analysis because the database contained no reliable information on their income. We imposed a maximum age of 44 for inclusion in the sample, as the probability of having no living parent rises with age. We assumed that setting an age limit of 44 would mean that for most couples in the sample, their parents would still be alive. We did not have any information on the status of the parents, as we only observed them in our panel when the adult children returned to the parental home. With these restrictions taken into account, we used data from 25,444 mothers, of whom 57.9% were divorced and 42.1% had been previously living with a partner without being married. We used data from 20,108 fathers, of whom 63.4% were divorced and 36.6% had experienced the dissolution of an unmarried cohabitation.
The data used for the analyses was organised as a person-file. Longitudinal information for each year from 2007-2013 was used to determine whether the parents returned to the parental home within 4 years (t + 4) of their break-up in 2007, 2008, or 2009. For the independent variables and the control variables, longitudinal information was used only from the year prior to the break-up (t−1). Since we were estimating person-level logistic regressions, no time-varying information was included in the analyses. The structure of the data is illustrated in Fig. 5.1.

Variables
The dependent variable was a dummy variable indicating that an individual had returned to the parental home after a break-up. When the administrative data indicated that the individual had registered at the same address as (at least) one of his/ her parents, the dependent variable showed unity. We did not take the timing of the return into account. Thus, whenever we observed a return, the dependent variable was coded as "return" (1). We had no information on who moved in with whom, but with the age selection set at 44, we assumed that the ex-partner joined the parental home, and not vice versa (for example, in response to care needs). Therefore, the dependent variable should be interpreted as merely referring to the ex-partner sharing a household with (at least) one parent. All models were controlled for the migrant status of each respondent. We distinguished the fathers and mothers who were of Belgian origin from those who were of Turkish or Moroccan origin, and were second-(or later-) generation migrants. The Belgian respondents were used as a reference category in all models. In addition to controlling for the background of each respondent, we controlled for the background of the former partner. We used three dummies representing (1) the first generation, (2) the second generation of the same origin, and (3) the mixed character of the couple when the partner was of another origin.
We controlled for a number of background variables. These variables refer to differences between the formerly married or cohabiting partners, welfare dependency, age (mean-centred plus age squared), and region (Flanders, Wallonia, Brussels Capital Region). The control variable "having young children in the household" (at t−1) measured the presence of children younger than 3 years old. The total number of children of the ex-couple is self-explanatory. Labour market participation was measured with two dummies: employed full-time (more than 80%) or parttime (80% or less). The reference category was made up of respondents who were either not active on the labour market or unemployed. Individual income included the respondent's earnings from employment, as well as any public transfers s/he received due to the disability or career interruption of the ex-partner. Childcare transfers or partner alimony payments were not included in the income data. Since partner alimony was structurally reduced by the law of 2007, only the absence of information on childcare transfers limited our ability to assess the respondents' total income. In the Belgian context, welfare benefits are based on an individual's prior labour market experience and household size. There are no specific public subsidies aimed at separated or divorced men and women or immigrants. However, previous studies have shown that people with a migrant background tend to have a higher level of welfare dependency than their native-born counterparts (Carpentier et al. 2014). All models were also estimated separately for men and women.

Analytical Strategy
The register data allowed us to determine for each respondent whether the boomerang effect was observable 4 years after the break-up, while controlling for preseparation characteristics. Using such a large sample had several advantages. First, the statistical power was greater, making the parameter estimates more robust. Second, it allowed us to examine the boomerang effect in more detail (e.g., including the dissolution of cohabitating unions among groups with a migrant background). As far as we know, no other study has ever combined a sample of postdissolution trajectories of this magnitude with a focus on respondents with a migrant background.
We used binary logistic regression to model the return to the parental home within 4 years of a union dissolution (Mortelmans 2010). Given that our follow-up period was limited to 4 years and our data were yearly, rather than quarterly or monthly, performing a discrete-time event history analysis would have offered little additional insight beyond that which a regular logistic regression approach could provide.
In a first step, we studied the effects of gender (M1) and migrant status (M2) for the complete sample by estimating models for both fathers and mothers. We then examined the interaction between gender and migrant status (M3).
In a second step, we estimated four models for fathers and mothers separately. The first two models introduced the migrant background (M1) and the partner background (M2). In Model 3, we added variables related to union type and economic status (income, dependency, and work status). Model 4 included the interaction between migrant background and socio-economic variables (employment, welfare dependency, and individual income). Model 5 included the interaction between migrant background and the characteristics of the former couple (union type and partner background).
Each model included all of the control variables. Table 5.1 provides the descriptive statistics by migration background (Belgian, Moroccan, Turkish). Belgians made up 86.9%, Moroccan migrants represented 9.5%, and Turkish migrants accounted for 3.6% of the sample. The composition of the couples in the sample that included a Belgian (ex-) partner was mostly homogeneous. Mixed couples made up less than 20% of the sample, and in most of these couples (16%), the migrant partner was from the second generation of any other origin group. It is important to note that in most of these previously mixed couples, the migrant partner had a migration background other than Moroccan or Turkish. Therefore, the frequencies in the column of Belgian couples did not add up to the frequencies of mixed couples in which a Moroccan or a Turkish migrant had a Belgian partner. Among the individuals with a Moroccan or a Turkish background, the shares of partners who were first or second generation were about the same. Mixed couples that included a Belgian partner represented around 10% of the sample, and there were almost no mixed couples that included a partner with another migrant background. Because the share of mixed (non-Belgian) couples was small, we decided to include both Belgian partners and partners and other nationalities in the reference category in our multivariate models. Looking at Table 5.1, we can clearly see that the comparison mainly involved Belgian partners. The individual characteristics of the ex-partners were generally similar. In terms of age, no major differences were found. However, the share of couples who were married was larger in the Turkish and Moroccan communities than in the Belgian sample. Although our aim was to include in the analysis equal shares of married and cohabiting couples (which we succeeded in doing in the original Belgian sample), the number of cohabitating couples among the Turks and the Moroccans was too low to allow us to include equal shares of the two relationship types. It also appears that the Moroccan ex-partners were more likely to have young children in the household than the Belgian or the Turkish ex-partners. Among the Belgian ex-partners, the regional spread was in line with that of the overall population: i.e., 60% were from the North (Flanders) and 40% were from the South or Brussels. Among the ex-partners with a migrant background, much higher percentages were from the Brussels Capital Region. When we examined the economic backgrounds of the former couples, we observed that large shares were working full-time before the break-up. However, compared to the Belgian couples, the couples in the Moroccan and Turkish communities were more likely to be inactive or unemployed, and to be dependent on welfare state transfers. These observations confirmed our assumption that the Moroccan and Turkish households were, on average, in a weaker economic position than their Belgian counterparts.

Regression Results
In a first step of the multivariate analyses, three models were estimated on the total sample (Table 5.2). The first model showed an overall effect that women with children returned less to the parental home after a break-up than men. This finding confirmed our hypothesis 1, which stated that the boomerang effect after a break-up is a gendered process. The second model, which added the migrant background of the parent, showed a significant positive effect of having a Turkish background. The Turks were more likely than the Belgians to return to the parental home. This was also the case for the Moroccans, but this effect was significant only after control variables were added. The effect for Moroccans was also much smaller than the effect for the Turks, and its significance was also on p < 0.05. Nevertheless, the model confirmed hypothesis 2, as it showed that Belgian ex-partners were significantly less likely to return to the parental home than ex-partners with a migrant background. When interacting gender and background in model 3 (hypothesis 3), we found that women in both groups were far less likely than men to return to the parental home. This was shown to be the case for both Turkish and Moroccan women; but again, the effect was found to be more pronounced in the Turkish community. To test our hypotheses by gender, we estimated separate models for men and women, as shown in Tables 5.3 and 5.4. Table 5.3 presents the results for men. In Model 1, we observed a strong positive effect of returning to the parental home for Turkish fathers. For Moroccan fathers, the effect was weaker, although still significant. When controlling for the previous partner (M2), we found higher odds of returning to the parental home for Moroccan fathers (M3). However, in the interaction models (M4 and M5), the effect of returning again disappeared for the Moroccan fathers. A second observation from Model 2 was that the migrant background of the previous partner mattered for men. When coupled with a partner from the first generation, men were less likely to return to the parental home. When the previous partner was from the second or third generation, men were no less likely to return to the parental home than they were if their former partner was Belgian. Having been married rather than cohabiting was found to slightly increase the likelihood of returning to the parental home.
Both of the income-related components were shown to be negative. In Model 3, we learned that having a higher income reduced the odds of returning to the parental home after a break-up. In addition, the more a father depended on welfare benefits, the less likely he was to return to the parental home. However, being employed increased the probability of returning for fathers, but only if they were working fulltime (as was the case for most of the fathers in the sample).
Models 4 and 5 in Table 5.3 tested the hypotheses on the interactions between migrant background and economic resources and family composition. With respect to economic resources, the findings indicated that having a high individual income had different implications for Belgians than for individuals with a migration background. Compared to their Belgian counterparts, higher-income individuals with a migration background were more likely to return to the parental home. For fathers with a Moroccan or -in particular -a Turkish background, the pattern even reversed, with the association between individual income and the likelihood of returning to the parental home turning positive. Moreover, compared to Belgian fathers, the effects of full-and part-time employment Moroccan and Turkish fathers were reversed, although these differences were not statistically significant. The effect of welfare dependency on the likelihood of returning to the parental home was found to be rather similar across all three groups. The interaction in Model 5 showed that the effect of whether a couple had been married on the likelihood of returning to the 3.12 * * * 3.21 * * * 3.31 * * * 3.36 * * * 3.30 * * * Source: Belgian Crossroads Bank for Social Security, calculations by the authors Notes: All models are also controlled for Age, Age 2 , region in Belgium, young child (<3 year) in the household; number of children in the household. Significance levels: * p < .05; * * p < .01; * * * p < .001 Full-time (t−1) 0.47 * * * 0.50 * * * 0.48 * * * Part-time (t−1) 0.29 * * * 0.33 * * * 0.30 * * * Welfare dependency (t−1) Turkish × Partner = 2nd generation 0.14 Source: Belgian Crossroads Bank for Social Security, calculations by the authors Notes: All models are also controlled for Age, Age 2 , region in Belgium, young child (<3 year) in the household; number of children in the household. Significance levels: * p < .05; * * p < .01; * * * p < .001 parental home did not vary by origin group, or by the migration background of the partner (i.e., no differences were observed between Moroccan and Turkish fathers). The main effect of having a first-generation ex-partner decreased in this model, but stayed significant. For women, the same five models were estimated (see Table 5.4). The first important result was found in Models 1 to 3: namely, that the main effects of the migration background of the woman and of her former partner on her likelihood of returning to the parental home were non-significant. We already showed in Table 5.2 that mothers were significantly less likely than fathers to return to the parental home. This overall pattern did not differ among Belgian, Moroccan, and Turkish mothers. When we looked at the effect of the woman's previous partner, we found that having a first-generation partner had a significant negative effect, but that this effect disappeared as soon as we controlled for other factors in Model 3. Again, our results indicated that the characteristics of a mother's previous partner did not play a role in her decision to return (or, rather, not to return) to the parental home. No differences in the likelihood of returning to the parental home were found between mothers depending on whether they had been married or cohabiting.
Individual economic resources played a similar role for all mothers. Among mothers, having a higher income or being dependent on welfare was associated with a lower likelihood of returning to the parental home. The effect of work status on the likelihood of returning to the parental home for mothers was similar to that for fathers. Mothers who were working full-time or part-time were more likely than those who were inactive or unemployed to return. It thus appears that the decision to return to the parental home was affected by the challenges single mothers face in combining work and family. Like for fathers, the interaction effects with migrant background (Models 4 and 5 in Table 5.4) for mothers did not differ much across the Moroccan and Turkish origin groups. The findings indicated that welfare dependency played a different role for Turkish than for Moroccan mothers, although why this was the case is unclear. The results further showed that Moroccan women who had been married were more likely than those who had been cohabiting to return to the parental home following a union dissolution.
Before turning to the discussion, we should point out the effects of some control variables (not shown in the tables). Age was found to be highly significant and negative both among men and women. The registers did not provide an indicator for home ownership or assets. However, age can be considered a proxy for this indicator, as older people are more likely than younger people to own a home and to have sizeable financial assets. Thus, it appears likely that the negative effect of age, in addition to the negative effect of income, confirmed our assumption that the accumulation of wealth gave the former partners more freedom, and enabled them to remain independent of their parents. The regional differences we observed were also striking, with the boomerang effect being more pronounced in the North (Flanders) than in the South (Wallonia) or Brussels. It is, however, possible, that this effect was conflated with the legal status of the former relationship, as marriage is still more common in Flanders than in Wallonia.
With regard to children, we saw no effects in any of the models for men. Among men, neither the presence of young children nor the total number of children appeared to played a role in the decision to return to the parental home. The findings showed that among women, both effects were highly significant, but were in opposite directions. The presence of young children made it more likely that a woman would to return to the parental home, which illustrates the importance of work-life balance for single mothers. However, effect of the total number of children was negative, which indicates that having a larger family made it more difficult for a woman to return to her parents' home.

Discussion and Conclusion
This research addressed several new issues related to the study of the "boomerang effect". We focused on the housing strategies of ex-partners with children after the end of either a marriage or a cohabiting relationship. Whereas most previous studies on adult children returning to the parental home focused on young adults, we focused on adults who had children of their own. Furthermore, we expanded our knowledge on this topic by looking at how cultural heterogeneity and gender influenced the likelihood of returning to the parental home. While the divorce rate is still lower among migrant couples with a Turkish or a Moroccan background than it is among native Belgians, it has been rising (Corijn and Lodewijckx 2009). Moreover, although the housing choices of individuals are subject to the same general mechanisms (e.g., their economic resources), regardless of their migration background, it is likely that the propensity of newly single parents to return to the home of their own parents is related to their cultural background and the societal position of their ethnic group.
Among people with a Turkish or Maghreb background -who make up 2.1% and 3.2% of the Belgian population, respectively (Noppe et al. 2018) -maintaining family and community ties and intergenerational support are often valued over asserting autonomy and independence from one's parents. However, in communities that hold on to these more traditional values, divorcees are likely to face stigma and negative attitudes, especially if they are women (Koelet et al. 2009b). Our expectation (hypothesis 3) that fathers with a Turkish or a Moroccan background would be more likely to return to the parental home than mothers of the same origin groups was only partly confirmed after controlling for relevant socio-economic and demographic factors. The initial models showed that, net of other effects, fathers were more likely than mothers to return to their parents' home (hypothesis 1), and that these associations were stronger for fathers and mothers with either a Turkish or a Moroccan background (hypothesis 2).
The separate models for men and women revealed that economic and family characteristics did not affect all of the subgroups in the same way. More specifically, when we interacted migrant background with these characteristics, we found that family and economic characteristics operated differently for men and women with migrant backgrounds. We found, for example, that for men with a Moroccan or a Turkish background, the interaction term for economic resources was both positive and larger in absolute terms than the negative base effect. In other words, while having a higher income decreased the probability of returning to the parental home for Belgian men, it increased probability for men with a migrant background. These results suggest that for these men, the cultural norm of maintaining close ties with family outweighed their desire to live independently. Another potential explanation for this finding is that these men faced discrimination in the housing market. A survey conducted by Heylen et al. (2007) found that 26% of Flemish owners of apartment buildings said they prefer to look for another tenant if an applicant has a migrant background. Heylen et al. (2007) found that although this share was lower among the private owners surveyed who rent out social housing, it was still 8.6%. These findings imply that whether it is due to actual or statistical discrimination, people with migrant backgrounds have, ceteris paribus, more trouble finding suitable housing than native Belgians, even when they have sufficient means to live independently.
No such associations were found for women. However, Turkish mothers who were dependent on welfare were shown to have a lower probability of returning to the parental home than other groups. This might be because these single mothers were at risk of losing (part of) their means-tested benefits if they moved into their parents' household. Another potential explanation for this finding is that because women tend to be the primary carers for their children, these women might have been eligible for social housing. Moreover, as the stigma attached to divorce is greater for women than for men in the Turkish and Moroccan communities (Koelet et al. 2009b), these women might have preferred to be independent. This would also explain why the same association was not found for men.
There are several limitations to our study. First, although it is known that children are more likely to reside with their mother than with their father after a divorce, we were not able to explicitly control for these arrangements. As a result, the associations we found might suffer from omitted variable bias. In addition, we could not control for the characteristics of the parents of the separated and divorced individuals in our sample. These characteristics are important, as they might have shown that in some of these families, returning home was not feasible economically, or that the parents' values either facilitated or hindered the return of their child after a divorce or a separation. Finally, the register data did not allow us to control for the educational attainment or the job status of either the adult children or their parents. Since these are crucial components of an individual's SES, this is a notable blind spot in our analyses. Even though the registers offer a large amount of statistical power, missing information on educational attainment or on the parents' characteristics (e.g., the size of the parents' residence) is a considerable limitation of this study. Higher levels of education are usually associated with more liberal attitudes towards marriage and divorce (Kalmijn and Kraaykamp 2007), although these effects are likely smaller for migrant populations than they are for majority populations (Kalmijn and Kraaykamp 2018).
In conclusion, this research has furthered our understanding of the tendency of parents to return to the home of their own parents after a relationship break-up by looking at differences in these patterns by gender and cultural background, and at the effects of socioeconomic and family indicators on the likelihood of returning. Research on divorce and separation has consistently shown that union dissolution is associated with financial downturns, especially for women. One of the strategies divorcees use to deal with their challenging circumstances is to return to the parental home. Especially when (young) children are living in the household, this strategy could free up parents to explore other potential coping strategies, such as increasing their labour market activities or finding a new partner. There are, however, barriers to employing these strategies, such as the welfare traps inherent to the system. Conversely, cultural norms can encourage or discourage a return to the parental home. Future research should investigate whether the decision to return to the parental home is beneficial for the financial and/or the subjective wellbeing of both the divorced parents and their children. Given that around 20% of Belgian children are living in households that are at risk of poverty (Vandenbroucke and Vinck 2015), gaining additional insight into the effectiveness of the various coping mechanisms parents turn to after a relationship ends is indispensible.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.