Please direct all correspondence to Monica J. Grant, 8128 Sewell Social Science Building, 1180 Observatory Drive, Madison, WI 53711, USA; or by ude.csiw.css@mtnarg
Copyright notice
The publisher's final edited version of this article is available free at Popul Stud [Camb]
Abstract
In this paper, we use longitudinal data to investigate how parental death and divorce influence young women’s own experience of divorce in Malawi, a setting where women marry relatively early and unions are fragile. We find that maternal death and parental divorce are positively associated with divorce for young women but, after controlling for socio-demographic and marital characteristics, only the association with maternal death remains statistically significant. Maternal and paternal death are both strongly associated with women’s post-divorce living arrangements, which in turn affects their material well-being. This finding suggests that divorcing at a young age shapes the subsequent life chances of women; although some women return to their parental home and may have the opportunity to reset the transition to adulthood, other women begin their twenties as head of their own household and with considerable material disadvantage.
Keywords: divorce, living arrangements, transition to adulthood, orphan, Malawi
Introduction
Studies of the transition to adulthood in sub-Saharan Africa have frequently examined the timing and sequencing of key role transitions, such as leaving school, sexual activity, marriage, and parenthood [Lloyd 2005; Mensch et al. 2005; Magadi and Agwanda 2009; Beguy et al. 2011]. Parental death and divorce have been found to accelerate these transitions, potentially compromising the life chances of African youth by limiting the material resources and family support that are available during the transition to adulthood [Tambashe and Shapiro 1996; Goldberg and Short 2012; Goldberg 2013; Marteleto et al. 2016]. However, research on the consequences of these dimensions of vulnerability for the well-being of young adults who have completed these role transitions has been limited. In this paper, we examine how parental death and divorce continue to influence the lives of young women after marriage and into early adulthood in Malawi, where the median age at first marriage is relatively young and partnerships are fragile.
Specifically, we use data from the Malawi Schooling and Adolescent Study [MSAS], a longitudinal survey in southern Malawi that followed a cohort of 14–16-year-olds from 2007 to 2013, to examine how parental death and divorce are associated with the likelihood that young women will divorce, as well as how these parental statuses are associated with the living arrangements and material well-being of young women after their union dissolves. Scholars have extensively examined and theorized how parental divorce and, to a lesser extent, death are associated with offspring divorce in western settings [Amato 2000; McLanahan and Percheski 2008; McLanahan et al. 2013] but, to our knowledge, no study has examined this question in sub-Saharan Africa. This lack of evidence exists despite reasons to believe that parental death and divorce may have different consequences for youth in sub-Saharan Africa, making this association of both theoretical and policy interest.
Divorce is common in our sample of young women: by the final survey round in 2013, when respondents were 20–23 years old, almost 35 per cent of ever-married women had divorced. This prevalence is consistent with other studies of divorce in Malawi [Reniers 2003; Bertrand-Dancereau and Clark 2016]. We find that both maternal death and parental divorce are positively associated with divorce for young women, although the association between parental divorce and a woman’s own divorce is no longer statistically significant after controlling for socio-demographic and marital characteristics. Maternal and paternal death—but not parental divorce—are both strongly associated with young women’s post-divorce circumstances, particularly in terms of who they live with after the divorce, which in turn is strongly associated with their material well-being. These findings illustrate the importance of considering the implications of parental divorce and orphanhood for young adults beyond the experience of the traditional adulthood markers.
Consequences of parental death and divorce
The consequences of parental death have been extensively studied in sub-Saharan Africa, motivated in part by concerns about rising orphan prevalence in the wake of the AIDS epidemic [Bicego et al. 2003; Grassly and Timaeus 2005]. Evidence suggests that although extended family systems are largely able to absorb orphans, the economic, emotional, and social capacity of families to provide care to orphans is not always sufficient [Abebe and Aase 2007]. The death of a parent may not only reduce household income and shift household labour demands, but also create affective dimensions of disadvantage, with a recent study finding that orphans in Lesotho received less ‘kindness and care’ from guardians relative to parents [Goldberg and Short 2012]. In addition, community and state support programmes for orphans run the risk of straining family relationships when orphans are perceived as privileged over other categories of vulnerable children, creating jealousy that may disadvantage orphaned youth within the extended family [Dahl 2014; Schenk 2009]. Resource discrimination within the family, combined with the psychological trauma of parental death and the loss of parental love, may, therefore, set young people on ‘trajectories of risk’ during the transition to adulthood [Mojola 2011].
Literature examining the consequences of parental death in sub-Saharan Africa has found evidence of orphan disadvantage in some, but not all, life course transitions. Researchers have consistently found evidence that being an orphan, especially as a result of maternal death, is associated with lower school participation and attainment relative to non-orphans [Case et al. 2004; Case and Ardington 2006; Evans and Miguel 2007; Beegle et al. 2010]. Orphans are also more likely than non-orphans to begin sexual activity at younger ages, engage in riskier sexual relationships, and give birth at younger ages [Birdthistle et al. 2008; Palermo and Peterman 2009; Mojola 2011; Chae 2013; Kidman and Anglewicz 2014]. In contrast, most studies have failed to find consistent evidence for an association between orphanhood and the timing of marriage [Beegle and Krutikova 2008; Palermo and Peterman 2009; Chae 2013; Grant and Soler-Hampejsek 2014]. Substantially less research has examined the consequences of parental divorce in sub-Saharan Africa. What literature does exist has focused on the consequences for children, finding evidence for higher child mortality, higher levels of child fostering, and lower grade attainment [Clark and Hamplova 2013; Grant and Yeatman 2014; Chae 2016]. No studies that we know of have examined how parental death or parental divorce are associated with divorce and its consequences for women in sub-Saharan Africa.
In contrast, the intergenerational consequences of divorce have been extensively examined in western settings [Amato 2000; McLanahan and Percheski 2008; McLanahan et al. 2013]. This large body of work has consistently found that children of divorced parents are themselves more likely to divorce [Amato 1996; Amato and DeBoer 2001; Teachman 2002; Wagner and Weiß 2006], though evidence suggests that this association is smaller in magnitude when divorce is more common, either over time or across countries [Dronkers and Harkonen 2008; Wolfinger 2011]. Given the relative rarity of parental death for children and young adults in the United States [US] and Europe today, fewer studies have examined the consequences of parental death for offspring divorce and none of the studies has differentiated the consequences of maternal and paternal death as we do. The handful of existing studies have found evidence suggesting that both parental death and divorce are associated with a higher risk of offspring divorce, although the associations tend to be weaker when parental absence is a result of death rather than divorce [ McLanahan and Bumpass 1988; Diekmann and Engelhardt 1999; Teachman 2002; Sigle-Rushton and McLanahan 2006; however, see Bumpass et al. 1991 and Bumpass and Sweet 1972 for exceptions].
There are reasons, however, to believe that parental death and divorce may have different consequences for youth in sub-Saharan Africa than for their counterparts in the US or Europe. First, the prevalence of parental death is much higher in many African countries than seen in the US or Europe as recently as the 1950s. Recent estimates from across a range of countries indicate that between 5 and 21 per cent of all children under the age of 15 have at least one deceased parent [Beegle et al. 2010]. This higher prevalence translates to different cultural framings of how orphans should be absorbed into the extended family. However, recent mortality patterns—particularly those driven by AIDS—as well as other stressors, such as limited land and food scarcity, are making kin less available and straining family resources in ways that may severely disadvantage youth as they become young adults [Blerk et al. 2008].
The transition to adulthood in most African countries also occurs at much younger ages than elsewhere. Although these transitions represent the assumption of adult responsibilities by young people, families help young adults navigate these new roles to the extent they are able. Support from parents and other relatives may provide women with greater decision-making power within the union, tools to enforce marital expectations and agreements, and a safety net to fall back on. Young women who marry as orphans or with divorced parents may have weaker access to these family support structures, either because kinship ties have been disrupted or strained, or because the resources of surviving parents are already stretched to their limits. African families are heterogeneous, varying widely by kinship structure, inheritance patterns, and expectations of family obligations [Oheneba-Sakyi and Takyi 2006]. It is important, therefore, to consider how parental death and divorce affect young adult outcomes in these settings.
Consequences of divorce for women
The literature on the consequences of divorce for women in sub-Saharan Africa is extremely sparse: scholars have tended to focus on the well-being of widows [Thomas 2008; Chapoto et al. 2011; van de Walle 2013] or the consequences of divorce for children [Clark and Hamplova 2013; Grant and Yeatman 2014; Chae 2016]. One exception is a recent study that used Demographic and Health Survey [DHS] data from 20 sub-Saharan African countries to examine the association between women’s marital status and their nutritional well-being [Djuikom and van de Walle 2018]. Overall, the study found that both widowed and divorced women tended to be worse off nutritionally, with lower BMIs, for example. In the case of Malawi, the negative association between being divorced and BMI was particularly strong in rural areas and persisted after controls for HIV status were included. In this paper, we build on these findings on the consequences of divorce by honing in on the case of Malawi, where women’s experience of divorce starts at a young age and occurs frequently. We use longitudinal data to examine women’s material well-being and living arrangements after divorce and how these circumstances are associated with whether their parents are alive and still married.
The more extensive literature from western contexts suggests that parental death and divorce may impair women’s material well-being after a union dissolution by disrupting access to the kinship networks that serve as a private safety net. Research from the US has found that women use three main strategies to lessen the economic blow of divorce: increasing their labour supply, obtaining cash assistance from government programmes, and drawing on their personal networks [Tach and Eads 2015]. Qualitative and quantitative studies have shown that assistance from family and friends serves as a lifeline for individuals in difficult economic times [Stack 1974; Edin and Lein 1997; Henly et al. 2005]. For low-income families, cash transfers across households may occur, but are often small due to the constrained economic resources of the people in their network [Edin and Lein 1997; Roschelle 1997; Jayakody 1998]. Instead, research has suggested that a major way low-income households provide support is through co-residence and the pooling of non-cash resources, such as childcare and food, which is often implicit in this type of living arrangement [Haider and McGarry 2005]. The first two strategies—increasing labour supply and receiving government support—might be challenging to access for low-income women in western contexts, but they are even less available in most countries in sub-Saharan Africa, where formal employment opportunities are limited [Filmer and Fox 2014; International Labour Organization 2015] and state social protection programmes remain nascent [Niño-Zarazúa et al. 2012]. In this context, support from personal networks is an important cushion for women experiencing divorce. Although the labour market and welfare state shape women’s options, women within a particular setting vary in the extent to which they can rely on a private safety net. Given that the informal safety net is an indication of social capital [Devereux 2001], women with weaker social ties are more vulnerable to economic shocks. In addition, parents’ situations affect their ability to provide assistance as well as the extent to which adult children are willing to ask for it.
Although family support is potentially of greater importance to recently divorced women in contexts such as Malawi, the strain on this coping strategy has grown as much of the population faces an increasingly stressful and resource-scarce environment, with smaller plot sizes, greater food scarcity, and increased mortality from HIV/AIDS [Peters 2006; Blerk et al. 2008]. Despite the increasing pressure on family support, intergenerational transfers remain significant in rural Malawi. Using data from three rural sites, Kohler et al. [2012] found that 49 per cent of women aged 20–60 years had received financial assistance from their parents in the past two years. An even greater proportion of women [66 per cent] had received non-financial assistance, such as help with collecting firewood and cooking, from their mothers. Mothers also had a greater number of adult children living with them, on average, compared with fathers. In sum, although the family network does not assure social and material support in this challenging context, intergenerational support, both monetary and otherwise, remains an important aspect of life in this high-poverty setting.
Finally, the generally matrilineal nature of our study area may have an impact on the consequences of divorce for young women. In matrilineal communities, obligations to provide support for orphaned youth are stronger for maternal kin than paternal kin, and the death of a father may be less likely to mobilize care from the extended family than the death of a mother. Indeed, one study found that the death of a mother was viewed as worse for children than the death of a father in matrilineal areas of Malawi, while in patrilineal areas the deaths of a mother or father were seen as equally consequential [Cook et al. 1998].
Parental death and divorce in Malawi
In this paper, we ask how parental death and divorce are associated with the experience of divorce for young women in Malawi, a country in south-east Africa where half the population lives under the national poverty line [United Nations Development Programme 2016] and 70 per cent of economically active individuals are subsistence farmers [National Statistical Office 2015]. Life table probabilities of divorce in Malawi calculated from longitudinal data suggest that in some regions as many as 40–65 per cent of all unions may end in divorce [Kaler 2001; Reniers 2003; Grant and Soler-Hampejsek 2014], with the highest probabilities of divorce in the southern region, the location of this study. This pattern of high divorce has been attributed to the fact that most ethnic groups in southern Malawi are matrilineal, with rights to children and inheritance determined by the mother’s kin group [Mitchell 1962; Reniers 2003; Grant and Yeatman 2014]. Family support for women may be stronger in these societies, weakening women’s reliance on marriage and leading to higher divorce rates [Reniers 2003; Takyi and Gyimah 2007]. Given this high prevalence of early union dissolution, Malawi is an ideal setting for considering the consequences of parental death and divorce on the experience of divorce among young women. Furthermore, with such a high percentage of young women experiencing divorce, it is important to understand how this experience is affecting their well-being and their transition to adulthood.
In addition to early divorce in Malawi, the level of orphanhood is also relatively high. According to Malawi’s 2008 Census, 12.4 per cent of children aged under 18 are orphans, defined as having experienced the death of at least one biological parent [National Statistical Office 2008], and almost two-thirds of these parental deaths are estimated to be related to AIDS [UNICEF 2015]. The majority of orphans are cared for within their extended family, with friends and relatives providing additional support to households containing orphans. Beyond the private safety net of the family, the public safety net in Malawi is relatively limited; only 25 per cent of orphaned children are estimated to live in communities that provide dedicated orphan support services, and only between one-fifth and one-quarter of households containing orphans receive free food or maize distribution [Kidman and Heymann 2009].
Given the lack of prior research on divorce among young women in rural sub-Saharan Africa, it is difficult to predict with certainty the direction of some of the associations between parents’ marital and survival status and young women’s own experience of divorce. For example, orphans and young women whose parents are divorced may be at greater risk of divorce than non-orphaned women whose parents are still married to each other, due to reduced access to family support to mediate marital conflicts. On the other hand, in line with prior research on divorce from matrilineal settings [Takyi and Broughton 2006; Takyi and Gyimah 2007], women with two married parents might be more likely to divorce, given that they are able to rely on greater parental support as an alternative to remaining in a bad union.
With regard to post-divorce outcomes, we expect women with divorced parents to be less likely to live with extended families than orphans, who might have lived with extended family before marriage. Such women may also be less likely to move in with a parent than women whose parents are still married to each other; financial resources may be more stretched among divorced parents and those who have remarried may be unable to accommodate daughters [Grant and Yeatman 2014]. Given these anticipated living arrangements and the fact that orphans cannot receive financial assistance from parents, we expect material well-being to be worse among women with divorced parents and orphaned women than for women whose parents are alive and married. In matrilineal settings, maternal orphans may be able to make stronger claims for support on maternal kin than paternal orphans. In addition, given limited economic opportunities for women relative to men in rural Malawi [Filmer and Fox 2014; Kendall and Silver 2014], the death of a father may lead to a greater loss of economic resources at the household level than the death of a mother. Therefore, we also put forward the hypothesis that paternal orphans will be more disadvantaged than maternal orphans.
Data and methods
Using longitudinal data from the MSAS, collected from 2007 to 2013, this paper explores the intergenerational nature of divorce, in terms of its transmission and consequences, in two districts in southern Malawi. The initial 2007 sample consisted of 1,764 students [875 girls and 889 boys] aged 14–17 years who were randomly selected from the enrolment rosters at 59 randomly selected primary schools in the Machinga and Balaka districts. The probability of a particular school being included was proportional to the size of its enrolment in 2006. An additional sample of 885 out-of-school adolescents [462 girls and 423 boys] was identified through key informants at the schools and within the school catchment villages. The out-of-school sample is comparable to out-of-school youth of a similar age in southern Malawi. For example, female, out-of-school MSAS respondents had completed 4.6 years of schooling on average, compared with 4.7 years for 14–16-year-old, out-of-school women in the DHS [NSO and ICF 2015]. In this analysis we focus on women, given that only 33 per cent of the male respondents were married and, of those married men, only 70 [16 per cent] had divorced before the final survey round.
Dependent variables
First, we examine the transition to divorce among ever-married female respondents. Each respondent contributes one person-year of exposure for each year of marriage. For example, a woman who was married in the first survey round and remained continuously married until the final survey round would contribute seven person-years to the analysis. Each person-year of marriage is coded as being still married or ending in divorce. Respondents no longer contribute person-years of observation after divorce, and are censored if they remain married at the time of last interview. Less than three per cent of female respondents were lost to follow-up without an observed marriage. An additional 4.1 per cent of women were lost to follow-up after marriage but before the final survey round and before a divorce was observed; these respondents contribute person-years to the analysis until they are censored. Women were significantly less likely to be lost to follow-up if their father had died, their parents had divorced, or they had experienced a premarital birth. Eighty-three per cent of the original sample of women were married by the final survey round, yielding 3,958 person-years of observation. All analyses are restricted to women in their first marriage [N=1,108].
Second, for women who divorce, we examine living arrangements and material well-being in the survey round following the divorce. Living arrangements are categorized as whether the respondent was living independently as the head of her own household, had remarried, was living with at least one parent, or living with other relatives. The household roster asked respondents to list ‘everyone who lives with you under the same household head, who shares a common pot at mealtimes, and lives in a common compound’. If a respondent identified herself as the head of her household but listed a biological parent in her household roster, she was coded as living with at least one parent. Likewise, a respondent who lived with others was only coded as living independently as head of her own household if all other household members were 15 years old or younger. Among respondents coded as living with other relatives, one-third lived with a grandparent, one-quarter with a sibling, and the remainder were distributed across other types of relatives, primarily matrilineal kin. Material well-being is based on household ownership of 18 durable goods and housing characteristics: mattress, sofa, table, chairs, paraffin glass lamp, television, radio, mobile phone, mosquito net, bicycle, motorcycle, car, boat/canoe, non-dirt flooring, toilet facilities, tin roof, cooking fuel source, and electricity. Material well-being was evaluated using principal component analysis [Vyas and Kumaranayake 2006]: the first principal component estimated from the data in the first survey round was used as a standard to construct the well-being index for all survey rounds. The analyses of post-divorce material well-being and living arrangements include all ever-divorced women who were interviewed in the survey round following their divorce [N=357].
Independent variables
Our key explanatory variables are the survival and marital statuses of the respondent’s parents. We include time-varying indicators [for each person-year] of whether the respondent’s mother or father was dead and whether the respondent’s parents were both alive but no longer married to each other. Preliminary analyses included the interaction between maternal and paternal orphanhood; these variables did not improve model fit and are not included in the final analyses.
In all models, we control for the respondent’s ethnicity, grade attainment, and whether she has had a child. The most common ethnic groups in the MSAS sample—Yao, Chewa, and Lomwe—are all matrilineal. Grade attainment is categorized as having less than primary, completed primary, or any secondary education. In the ‘transition to divorce’ models, the indicator of whether the respondent has had a child is time varying. These models also control for quartiles of material well-being measured at each person-year and characteristics of the marriage, including age at first marriage, age difference between the spouses, and union formality. In our study area, all cohabiting unions are referred to as marriages, although they may vary in the extent to which the union has been formalized. The most basic formalization involves the mediation of the ankhoswe, representatives from the families of the husband and wife who negotiate the terms of the marriage. Marriages may be further solemnized through a traditional or religious ceremony. Non-formalized cohabiting unions are often the product of elopements and have relatively low levels of family support [Bertrand-Dansereau and Clark 2016].
In the analyses of post-divorce outcomes, we control for the respondent’s age at divorce. Finally, in our analysis of post-divorce material well-being specifically, we also control for the respondent’s living arrangements, as previously defined. In the ‘transition to divorce’ models, all independent variables are lagged one year before the divorce outcome is measured. For respondents who were married at the first survey round, we use the values of time-varying variables measured at the first survey round. Similarly, for the analysis of post-divorce outcomes, all time-varying indicators are measured in the survey round immediately before the divorce.
Analysis
We use discrete-time logit regressions to examine the transition to divorce, such that:
log[Pit∕[1−Pit]]=α+βMumDeadit+βDadDeadit+βParDivit+βXit+βMit
where the discrete-time hazard of divorce is evaluated for each person-year of marriage, t, contributed by respondent i. The full model contains time-varying indicators of parental death and divorce, plus vectors of socio-demographic variables [X] and marital characteristics [M]. Preliminary analyses investigated the possibility that selection into marriage biased the analysis of divorce. In an alternate specification, we estimated two separate regressions: for whether women had married by 2011 and whether they had divorced within the first two years of marriage. We then predicted the residuals of these two equations and found that the errors were uncorrelated [R = −0.0171]. This finding suggests that selection into marriage is not biasing our analyses; therefore, we only present the results of the discrete-time event history analysis. Post-divorce living arrangements are analysed with a multinomial logit regression, such that:
ηij=[πij∕πiJ]=αj+βjMumDeadi+βjDadDeadi+βjParDivi+βjXi+βjMi
for each category, j, of post-divorce living arrangements, J. For ease of interpretation, we present the average marginal effects estimated from this model. Post-divorce material well-being is evaluated with a linear regression:
Yi=α+βMumDeadi+βDadDeadi+βParDivi+βXi+βMi
Although the full discrete-time regression includes all marital characteristics variables, the vector of marital characteristics [M] included in the post-divorce analyses is restricted to age at divorce and whether the respondent had given birth to a child by the survey round before the divorce. The analysis of living arrangements includes a lagged measure of material well-being, while the analysis of material well-being also controls for post-divorce living arrangements. Robust standard errors are clustered by sample school in all analyses. Finally, all descriptive statistics are weighted and adjusted for the sample design.
Results
Table 1 presents the descriptive statistics for the analytic samples. Of ever-married women, more than 32 per cent had divorced by the final survey round. In the first year of marriage, almost 22 per cent of young women had divorced parents, 15 per cent were maternal orphans, and 25 per cent were paternal orphans. Ever-married women in the analytic sample were more likely than women in the full sample of all female respondents to have divorced parents, but there were no significant differences in the prevalence of orphanhood [results not shown]. Given that the average age of first marriage for women who married by 2013 was only 17.3 years, these distributions suggest that the majority of women enter the transition to adulthood with constrained family support. However, it is also important to note that because this average age of marriage does not include women who marry at older ages, this figure is an underestimate of the average age of marriage in the study area.
Table 1
Descriptive statistics of analytic samples: ever-married and ever-divorced women in Malawi, 2007–13
Ever-married women [N=1,108]Ever-divorced women [N=357]Divorced by 201331.8100.00Living arrangements, post-divorcea Head of household–19.61 Remarried–27.45 Living with parent[s]–39.22 Living with other relatives–13.73Material well-being, post-divorce [s.d.]a–−0.30 [2.07]Independent variablesParents divorcedb21.825.6Mother deadb15.416.5Father deadb24.923.2Highest grade attained [s.d.]5.81 [1.87]5.56 [1.94] Less than primary54.872.4 Completed primary23.216.5 Any secondary21.911.1Material well-being [s.d.]b−0.04 [2.24]−0.13 [2.23]Material well-being, pre-divorce [s.d.]a−0.19 [2.00]Ethnic group Yao46.649.8 Chewa21.122.0 Lomwe21.118.2 Other11.110.1Had a childb14.219.3Marriage formality Elopement7.611.9 Ankhoswe only23.032.9 Traditional ceremony34.933.3 Religious ceremony12.910.9 Missing21.611.0Spousal age difference Same age or husband younger5.26.3 Husband less than five years older57.655.9 Husband five or more years older36.137.6 Missing1.10.2Age at marriage [s.d.]17.3 [2.0]16.7 [1.8]Age at divorce [s.d.]–18.69 [1.65]
Open in a separate window
aThese variables are restricted to ever-divorced respondents.
bThe distribution of these variables is presented for the first year of marriage.
Notes: All distributions are weighted and adjusted for the complex survey design. S.d. refers to the standard deviation.
Source: MSAS 2007–13.
Among ever-divorced women, only 20 per cent were living on their own in the survey round after divorce. In contrast, 27 per cent of women had already remarried. The most common arrangement [39 per cent] was to live with at least one parent. The remaining 14 per cent lived with other relatives. Following divorce, women’s mean material well-being was 0.30 standard deviations lower than the full MSAS sample mean at the first survey round. This post-divorce index score was, on average, 0.10 standard deviations lower than women’s mean material well-being index score during the survey round before their union dissolution.
Divorce
Table 2 presents the results of the discrete-time logit regression of the transition to divorce. Model 1 examines the association with parental death and divorce, controlling for years of marriage. Model 2 adds socio-demographic variables, while Model 3 adds marital characteristics. In the first model, women whose mother had died had 35 per cent [β = 0.31] higher odds of divorcing than women whose mother was still alive, while the odds of divorce were 24 per cent [β = 0.23] higher for women whose parents were divorced, although the latter association was only marginally significant. None of the models showed statistically meaningful associations between paternal death and the odds of divorce. When socio-demographic variables were included, in the second model, the association between respondent divorce and parental divorce became statistically non-significant. This attenuation is driven by lower educational attainment among women whose parents had divorced, consistent with prior research [Chae 2016]. The significant association between respondent divorce and maternal death, however, persisted after the addition of socio-demographic and marital characteristics to the regression.
Table 2
Transition to divorce, discrete-time logit regression for women in Malawi, 2007–13
Model 1Model 2Model 3βSEβSEβSEParents divorced0.230.12+0.180.120.170.13Mother dead0.310.14*0.300.15*0.330.14*Father dead0.020.130.020.140.030.14Years married−0.220.04***−0.240.03***−0.330.04***Highest grade attained [ref. less than primary] Completed primary––−0.530.16***−0.350.17* Any secondary––−0.790.20***−0.430.23+Material well-being [ref. low] Lower-middle––0.020.12−0.030.13 Higher-middle––−0.060.15−0.030.15 High––0.060.160.120.17Ethnicity [ref. Yao] Chewa––−0.080.14−0.060.13 Lomwe––−0.200.15−0.140.16 Other groups––0.020.160.080.16Age at marriage––––−0.190.04***Had a child––––0.510.13***Union formality [ref. traditional ceremony] Elopement––––0.500.12*** Ankhoswe only––––0.640.21** Religious ceremony––––−0.280.20 Missing––––0.050.20Spouse age difference [ref. Husband less than five years older] Same age or husband younger––––0.780.28** Husband five or more years older––––−0.130.12 Missing––––−1.491.08Constant−1.770.11***−1.460.16***1.360.61*Number of respondents1,1081,1081,108Person-years of exposure3,9583,9583,958Log pseudo-likelihood−1261.95−1245.11−1201.64Wald chi248.94***98.71***210.82***
Open in a separate window
+p