Number 35: Marriage breakdown in Australia: social correlates, gender and initiator status

Executive summary

This paper presents the main findings of an Australian Research Council linkage project with the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and the University of Queensland. One of the main aims of the project was to obtain a better understanding of marriage breakdown in the Australian context, with a particular focus on gender differences in the social correlates of marriage breakdown. In this paper I present some of the key findings of that project. The project makes two main contributions to previous research into marriage breakdown in Australia: first, it examines gender differences in the social determinants of marriage breakdown in Australia; and second, previous research is extended by examining whether there are gender differences in the decision to separate, taking into account social characteristics.

A large body of literature identifies a range of social characteristics—including temporal and family background factors, relationship and fertility histories, attitudes and beliefs, and socioeconomic position—that are associated with marriage breakdown. Understanding how these factors are differently associated with marriage breakdown for men and women provides a better understanding of why some marriages break down and others remain intact. In this paper I employ two different approaches to examining gender differences in the social determinants of marriage breakdown. First, I examine gender differences in the associations between social characteristics and marriage breakdown. Second, I examine gender differences in the associations between social characteristics and which spouse initiated separation.

The analysis uses retrospective data from all persons who were currently or had been previously married at Wave 1 of the Household Income and Labour Dynamics in Australia (HILDA) survey (2001). The study is limited to first marriages and excludes the marriages of migrants in the sample that had ended prior to their migration to Australia. The final analytic sample consists of 8,993 first marriages: 4,110 men and 4,883 women. The findings can be grouped into three main themes.

Social characteristics are important for understanding marriage breakdown

  • Men and women in older birth cohorts are less likely to divorce than those in younger cohorts.
  • In Australia, migrant women from non-English speaking backgrounds have a lower risk of marriage breakdown than Australian-born women. Migrant men from English-speaking backgrounds have a lower risk of marriage breakdown than Australian-born men. Migrant women from English-speaking backgrounds have a greater risk of marriage breakdown than migrant men from English-speaking backgrounds.
  • Having divorced parents, living together before marriage, having children before marriage or in the first year of marriage and marrying young all increase the risk of marriage breakdown. For women, young age at marriage was associated with a greater increase in the risk of marriage breakdown compared to men.
  • Higher levels of religiosity and having children in marriage reduce the risk of marriage breakdown.
  • Higher levels of education increase the risk of marriage breakdown for women, while in contrast, higher levels of education decrease the risk of marriage breakdown for men.

Women are more likely to initiate separation than men

  • Thirty per cent of separations were initiated jointly with 70 per cent of separations initiated unilaterally by either the husband or wife. Of these unilateral separations, 69 per cent were wife initiated and 31 per cent were husband initiated.

Gender differences in the decision to separate or remain married

  • Women are more likely to initiate separation than comparable men when they migrate from English-speaking countries, when they marry young and when they have higher levels of education.
  • Men with higher levels of attachment to religion are less likely to initiate separation than comparable women.
  • The main gender difference observed is that, according to women's reports, very few women's social characteristics are associated with husbands initiating separation. This patterning is not evident for men, whose characteristics are more evenly associated with their reports of both husbands and wives initiating separation.

[ Return to Top   Return to Section ]

1. Marriage breakdown in Australia

As in most western nations, marriage and family life have undergone major changes in Australia since the end of World War II. Overall, rates of marriage have declined (de Vaus 2004); the number of couples living together before, or instead of, marriage has increased dramatically (Glezer 1997); the number of children born in de facto unions rather than marital unions is increasing (de Vaus 2004); women are delaying child-bearing and couples are having smaller families (McDonald 2000). Arguably though, the change that has had the most wide-ranging and far-reaching consequences for family life is the increase in divorce. While the increase in the rate of divorce in Australia has slowed since the 1980s—and may have stabilised—in the year 2000 more marriages dissolved due to divorce than widowhood for the first time. Divorce continues to be a pervasive feature of Australian social life: 32 per cent of current marriages are expected to end in divorce and it has been predicted that this may increase to 45 per cent over the next few decades if current trends in recent marriage cohorts continue (Carmichael, Webster & McDonald 1996).

Many, although not necessarily all, couples whose marriages break down are able to successfully move on with their lives, and those with children often renegotiate their post-divorce relationship in positive ways (Funder 1996; Smart 2000; Smart & Neale 1999). Nevertheless, there are major social, emotional and financial implications for separating and divorcing couples (Amato 2000; de Vaus 2002; Smyth & Weston 2000). Further, there are major changes in the conduct of family life after separation and divorce. Many children lose regular face-to-face contact (and some lose contact altogether) with their non-resident parent, who is usually the father (Smyth 2005; Smyth, Caruna & Ferro 2005). Co-parenting, re-partnering and the formation of step and blended families after divorce have added to the diversity of family and household forms in Australia, and the complexity of family and household arrangements (Beck-Gernsheim 2002; Cherlin 1978; Smart & Neale 1999). Whereas the death of a spouse represents a 'clean' break from the marriage, in the event of separation and divorce, there is a continued 'post-marriage marriage' involving the negotiation of shared parental responsibilities and parenting time schedules for children across households (Beck-Gernsheim 2002; Smart & Neale 1999).

The costs of marriage breakdown, however, extend beyond those incurred by the individuals and families involved. In 1994–95, divorce was estimated to cost the Australian community $3 billion in direct costs per year (House of Representatives Standing Committee on Legal and Constitutional Affairs 1998).1 Further, given that this estimate is now more than 10 years old, it is likely that this amount has increased. In addition there are indirect costs, which include (but are not limited to) hidden costs to the healthcare system because people of separated or divorced marital status typically have worse physical and mental health than those who are either currently married or have never married (de Vaus 2002; Lillard & Waite 1995). Absenteeism and low work productivity have also been linked to relationship problems (House of Representatives Standing Committee on Legal and Constitutional Affairs 1998).

1.1 The current marriage and divorce policy climate

Over the last decade divorce policy has concentrated on two broad areas of reform. The first is legal reform under the Family Law Act 1975, designed to reduce the emotional and financial burden of divorce. The second major focus of recent policy reform, and the most relevant for the purposes of this research, has been on policies that aim to prevent marriage breakdown. The parliamentary report To have and to hold: strategies to strengthen marriage and relationships highlighted the apparent failure of the court system and the Family Law Act in its operational form to provide adequate support to marriages and families (House of Representatives Standing Committee on Legal and Constitutional Affairs 1998). Since then, a number of government policy initiatives have focused on preventing separation and divorce, primarily by increasing funding to relationship and premarital education and marriage counselling services, and the evaluation of those services. Such initiatives include the Men and Family Relationships Initiative (1999–2000) and the Stronger Families and Communities Strategy (2000–2009). For example, under the Men and Family Relationships Initiative, funding was awarded to 46 community-based organisations to target men for counselling and educational programs designed specifically to assist them to deal with relationship difficulties (O'Brien & Rich 2003).

Most recently, the Australian Government allocated $397.2 million between 2005 and 2009 to implement changes to the family law system, including, but not limited to, expanding relationship support services and facilities to assist families. The services directed towards preventing marriage breakdown are primarily focused on marriage education and relationship counselling. Research suggests that these types of services are effective in reducing marital conflict and marriage breakdown across a range of social and demographic groups but have some limitations (Halford et al. 2006; Halford & Simons 2005; Stanley et al. 2006). The risk factors for separation and divorce are complex and extend far beyond the couple and individual dynamics on which these policy initiatives are focused. Halford et al. (2006) found that those couples who were most at risk of marriage breakdown were also the least likely to use marriage education or marriage counselling services.

While for some couples divorce is not necessarily a negative outcome and is the best solution to an unhappy or unworkable marriage, for many marriages that end in divorce it is believed that with appropriate help couples could have negotiated happy and workable marital relationships (Halford & Simons 2005). Research identifying the broader social and demographic factors associated with marriage breakdown and who initiates separation, such as that proposed here, can complement recent policy initiatives and aid in the identification of 'at risk' groups that can be targeted in prevention campaigns promoting the use of premarital education and relationship counselling (Halford 2000; Halford & Simons 2005; House of Representatives Standing Committee on Legal and Constitutional Affairs 1998).

1.2 Previous Australian divorce research and the contributions of this project

Internationally, a large and growing body of research has emerged to try and explain why some marriages end while others remain intact. Contemporary divorce research is dominated by two main approaches. One focuses on psychosocial explanations for marriage breakdown, or microrelational processes, such as couple behaviours and interactions (for example, Gottman & Notarius 2002) and people's perceptions of relationship quality and satisfaction (for example, Wolcott & Hughes 1999). The other approach examines sociostructural determinants of marriage breakdown, encompassing sociological, demographic, economic and life course predictors of divorce (for example, Bracher et al. 1993; Bradbury & Norris 2005).

The strength of the psychosocial approach is that it generates an understanding of the processes of marriage breakdown. However, it often relies on small, non-representative samples of separated and divorced couples and provides little insight into the contexts of marriage breakdown. In contrast, the sociostructural approach typically uses nationally representative samples and provides insight into the contexts and the circumstances within which marriage is conducted and the correlates of marriage breakdown. However, it offers little insight into the processes of marriage breakdown. Only a handful of studies combine both approaches, usually by including measures of both micro and macro predictors in models and examining the relative associations among them (see, for example, Amato & Previti 2003; Sanchez & Gager 2000).

A comparatively small but important body of research on marriage breakdown has emerged in Australia over the last few decades. It encompasses both micro and macro approaches. This Australian literature, however, is dominated by descriptive studies of the demographic trends in divorce (Carmichael, Webster & McDonald 1996; de Vaus 1997, 2004; Ozdowski & Hattie 1981; Stewart & Harrison 1982) and research into post-separation and post-divorce processes (Funder 1996; Funder, Weston & Harrison 1993; McDonald & Institute of Family Studies (Australia) 1986; Smyth 2005; Smyth, Caruna & Ferro 2005; Smyth & Weston 2000). This latter body of work grew exponentially with the interest in parenting arrangements after divorce following the 2004 parliamentary inquiry Every picture tells a story: report on the inquiry into child custody arrangements in the event of separation (House of Representatives Standing Committee on Family and Community Affairs 2003). Overall, this research provides little explanatory information about the predictors of marriage breakdown in Australia. Previous research indicates that more often than not the decision to separate is made by one spouse, without the knowledge or agreement of the other.

A handful of Australian studies specifically investigate social factors that increase or decrease the risk of marriage breakdown. For example, Wolcott and Hughes (1999) investigated separated and divorced people's perceptions of why their marriages failed. Though somewhat dated now, Burns (1981, 1984) also examined marital breakdown by investigating the reasons respondents gave for their separation/divorce. However, these studies are limited because they sampled separated and divorced people only and therefore provide no comparison of characteristics with people who remained married. Moreover, since the emphasis of the research was on the respondent's perception of why the relationship ended, neither study was able to fully investigate the broader social correlates of marriage breakdown.

Several other Australian studies have examined the sociostructural predictors of marriage breakdown (Bracher et al. 1993; Bradbury & Norris 2005; Day 1964; de Vaus, Qu & Weston 2003; Jones 1994; Roden 1989; Sarantakos 1994), but all have limitations. Jones (1994) compared divorce rates of mixed-ethnic marriages with ethnically homogenous marriages. Similarly, Khoo and Zhao (2001) investigated migrant divorce rates in Australia, adjusting for age and marriage patterns. And Day (1964) compared patterns and characteristics of divorce in Australia with those in the United States. Each of these studies used Australian Bureau of Statistics data for the divorced population only, with no comparison with those who remained married. Moreover, official data exclude those who were separated but not divorced, and therefore do not allow a comprehensive investigation of people whose marriages ended. Both Bracher et al. (1993) and Roden (1989) investigated the temporal and life course determinants of divorce in Australia, but only included women in their sample. The study by Bradbury and Norris (2005) was the first published longitudinal examination of socioeconomic determinants of marriage breakdown in Australia. It included both husbands' and wives' characteristics, but gave little attention to gender differences or the relative contributions of spouses' characteristics. A study by de Vaus, Qu and Weston (2003) examined the effects of premarital cohabitation on the likelihood of marriage breakdown within eight years of marriage; however, it did not specifically examine other social characteristics (even though some were included in the models as controls). Similarly, Sarantakos (1994) focused on the impact of cohabitation on marital quality and marriage dissolution and paid little attention to other factors. In sum, while previous research provides important background information and raises key issues, there is still a large gap in our understanding of gender differences in the relationship between sociostructural factors and marriage breakdown in Australia.

In this paper I make two main contributions to previous research: first, I undertake a more comprehensive investigation of gender differences in the social determinants of marriage breakdown in Australia than previous Australian research; and second, I extend previous research and examine whether there are gender differences in the decision to separate that are based on social characteristics. This study complements and extends previous Australian research on marriage breakdown in several respects. It uses a large, nationally representative population sample rather than purposive samples of separated and divorced people. I focus on structural and demographic determinants of marriage breakdown rather than respondents' perceptions of why the relationship broke down, or general divorce trends, which is where much current Australian divorce research is concentrated. In addition, I consider a broad range of social characteristics. Finally, I compare and contrast the associations between the social correlates of marriage breakdown for men and women, whereas previous Australian research has largely ignored gender differences in the social determinants of marriage breakdown.

Additionally, I build on the international literature on marriage breakdown and examine whether sociostructural predictors of marriage breakdown are associated with which spouse made the final decision to separate. With the exception of only three studies (England, Sayer & Allison 2005; Hewitt, Western & Baxter 2006; Kalmijn & Poortman 2006), all previous Australian and international research on the social predictors of marriage breakdown examines gender differences in marriage breakdown in the effects of husbands' and wives', or men's and women's, social characteristics on marriage breakdown. In contrast, I take into account gender in the dependent variable by examining which spouse initiated marital separation with three possible outcomes: wife-initiated, husband-initiated and jointly initiated separation. Therefore, rather than treating marriage breakdown as a uniform 'event', I differentiate it into three separate outcomes—a wife-initiated marriage breakdown, a husband-initiated marriage breakdown and a jointly initiated marriage breakdown.

This project is limited to examining legal marriage, which means that I do not examine relationship dissolution in de facto unions. While long-term de facto partnerships are growing in popularity and increasing numbers of children are being born in de facto relationships (de Vaus 2004), cohabitation remains a step into marriage and the majority of people in de facto relationships go on to legally marry. Moreover, marriage remains the dominant form of long-term partnering in Australia (Baxter 2003). Further, by only examining separation and divorce, I do not provide a comprehensive investigation of the stability or instability of marital relationships. There are other factors, such as quality of the relationship and relationship satisfaction, that are also important indicators of marital stability. Finally, the focus is on sociostructural factors, which have been found to be important predictors of marriage breakdown but do not offer much insight into the micro relational processes of marriage breakdown within couples, such as people's reported reasons for divorcing or the quality of their interpersonal relationships.

In the next section of this paper, I present an overview of the literature on the social correlates of marriage breakdown and initiator status (that is, which spouse initiates separation). In Section 3, I describe the data, analytic sample, measures and analytic approach used for the paper. In Section 4, I present the findings of the analysis examining gender differences in the relationships between selected social characteristics and marriage breakdown. In Section 5, I present the findings in relation to which spouse initiated the separation. Finally in Section 6, I provide conclusions and discuss the policy implications of the paper.

[ Return to Top   Return to Section ]

2. Social characteristics associated with marriage breakdown

A large and growing body of literature has identified a wide range of social factors that are associated with marriage breakdown, including temporal, life course, attitudinal and economic factors. Temporal factors reflect the influence of historical time, such as the social contexts and broad social trends that impact on marital breakdown (Finlay 2005; Thornton & Rodgers 1987). Life course factors relate to the timing of specific life events that precede marital breakdown. The characteristics commonly used in studies taking a life course approach can be grouped into two main categories, those relating to family background and those measuring relationship and fertility histories (White 1990). Some attitudinal factors, such as attitudes towards gender roles and religious beliefs, have also been found to be associated with marriage breakdown. Finally, socioeconomic factors have been found to be associated with whether a marriage continues or ends.

2.1 Temporal and family background

A life course conceptual framework operates from the position that social processes interact with time-dependent processes. Prior life experiences that stem from temporal and family background factors can influence marriage breakdown. From this perspective, some risk factors for divorce occur even before marriage partners are selected and marriage vows are taken.

Temporal correlates of marriage breakdown

Temporal divorce risk factors refer to the contexts and broad social trends that impact on divorce. It is well established that recent generations are more likely to divorce than older generations (Bracher et al. 1993; Heaton 1991). Some social changes that have occurred in Australia over the last century that are likely to contribute to differences between older and younger generations in their willingness to divorce include an increasing acceptance of divorce (de Vaus 2004); legislative changes that have made divorce a less costly and time-consuming process (Finlay 2005); a substantial increase in married women's participation in the workforce (Baxter 2003); and a shift in gender roles and the social expectations of marriage (Beck-Gernsheim 2002; Giddens 1992).

Several measures are used in the literature to capture the influence of historical time on divorce including age, marriage cohort and birth cohort (Bracher et al. 1993; Heaton 1991; Thornton & Rodgers 1987). The most commonly used measure is the respondent's birth cohort, which captures the combined effects of historical context (that is, morals, values, beliefs and economic conditions) during formative childhood years and ageing (Heaton 1991). Previous research has found that birth cohort has a curvilinear association with divorce, with older cohorts less likely to divorce than younger cohorts, but younger cohorts having shorter marriage durations and a lower risk of divorce than older cohorts. Typically, cohorts in the middle of the age distribution have the highest rates of marriage breakdown (Bracher et al. 1993).

Ethnicity

In Australia, ethnic origin has been found to be associated with marriage breakdown (Bracher et al. 1993; de Vaus 1997; Jones 1994; Roden 1989). This body of work suggests that compared to the Australian-born population, migrants from English-speaking backgrounds have higher rates of marriage breakdown while non-English speaking migrants have a lower risk of marriage breakdown. It is likely that the practices of marriage and family life in migrant populations differ from the mainstream population because they retain the cultural and religious beliefs and practices of their country of origin (Ware 1975).

Further, while specific ethnic groups are likely to experience different rates of divorce compared to Australians (and other ethnic groups), marriages that cross ethnic boundaries are particularly vulnerable (de Vaus 2004; Jones 1994). Jones (1994) argues that a higher rate of divorce in cross-ethnic marriages is likely due to cultural differences between husband and wife in their expectations of the institution of marriage and how those differences are to be negotiated.

Family background

Family socioeconomic background is hypothesised to influence marriage breakdown in two ways. First, people from more prosperous and educated family backgrounds may have more stable marital histories because they have been exposed to less hardship and turmoil while growing up (Wolfinger 1999). Alternatively, parents with higher levels of education are more likely to impart social and cultural capital to their children, which increases children's marital stability (Wolfinger 2000). In particular, there is an expectation that children with more economically stable family backgrounds will have had better childhood experiences and better role modelling and therefore be better positioned to have stable marriages.

Findings of previous US research linking family socioeconomic background and divorce are mixed. For instance, Bumpass et al. (1991) found that women whose mothers had lower levels of education were more likely to experience marital disruption, whereas Wolfinger (1999) found that respondents whose parents had not completed high school were less likely to divorce. In Europe, the results more consistently show an increased risk of divorce among respondents whose parents have higher socioeconomic status. A recent Dutch study found a strong positive association between parents' socioeconomic status, encompassing mother's and father's education and father's occupation, and marriage breakdown (De Graaf & Kalmijn 2006). Similarly, Lyngstad (2006) found in Norway that couples with highly educated parents were more likely to experience marriage breakdown than couples with less educated parents. These researchers suggest that rather than promoting more stable relationships, higher parental socioeconomic position may indicate more liberal views towards divorce, which could explain the increase in the risk of marriage breakdown. There is little evidence linking family socioeconomic background and divorce in the Australian context. Research by Bracher et al. (1993) found no association between mother's or father's education, father's occupation and marriage breakdown for women. The findings of this Australian study suggest that parental socioeconomic background may not be important for understanding marriage breakdown in the Australian context, although more needs to be done to confirm this finding.

The most consistent family background risk factor for marriage breakdown is parental divorce: respondents whose parents have divorced are more divorce prone. This phenomenon has been labelled the intergenerational transmission of divorce (Amato 1996; Beck-Gernsheim 2002; McLanahan & Bumpass 1988). The most common explanation is that children of divorced parents receive poor socialisation and role modelling for marriage (Kiernan & Cherlin 1999; Mueller & Pope 1977; Teachman 2002a; Teachman 2004). Amato (1996), using longitudinal data from the United States, found that the main causal mechanism for the intergenerational transmission of divorce was the increased likelihood that children of divorce exhibit behaviours such as jealousy that interfere with the maintenance of mutually rewarding intimate relationships. There is some Australian evidence in support of this explanation. In a longitudinal Australian study, Burns and Dunlop (2000) examined the extent to which personal qualities of children of divorced parents reported by parents and the children themselves predicted the quality of their early adult relationships. They found that children of divorced parents had more behavioural problems than children of intact families, which in turn negatively affected the quality of their intimate relationships 10 years later.

2.2 Relationship and fertility factors

A person's relationship and fertility pathways can also influence whether their marriage remains intact or ends in separation and divorce. Factors such as cohabiting before marriage, having children before marriage, having children during marriage and age at marriage are all associated with marriage breakdown in various ways.

De facto cohabitation

A popular belief exists that cohabitation improves the process of mate selection by giving couples the opportunity to get to know each other better, to negotiate roles and to develop communication skills prior to marriage (Becker 1981; Beck-Gernsheim 2002; Sarantakos 1994). There is, however, very little empirical support for the expectation that cohabitation reduces the risk of marriage breakdown. In Australia (Bracher et al. 1993; de Vaus, Qu & Weston 2005) and other comparable western countries such as the United States (Bumpass, Martin & Sweet 1991; DeMaris & Rao 1992; Dush, Cohan & Amato 2003; Teachman & Polonko 1990), Canada (Hall 1996; Hall & Zhao 1995) and Britain (Berrington & Diamond 2000), research consistently finds that people who cohabit prior to marriage have an increased likelihood of marriage breakdown compared to those who do not cohabit.

There are three main explanations for this finding in the research literature. The predominant explanation for this paradox is that there is a selection effect that operates in relation to cohabitation. The argument is that those who cohabit are less conventional in their attitudes towards relationships and marriage and have lower levels of commitment to marriage than those who do not cohabit. People who cohabit before marriage are therefore more likely to divorce (Hall 1996; Lillard, Brien & Waite 1995). Cohabitation prior to marriage thus reflects values and beliefs that are likely to increase the risk of marriage breakdown. Some studies have found that the cohabitation effect on divorce may be lessening in recent generations, for whom cohabitation has become more widely practised (de Vaus, Qu & Weston 2003; Schoen 1992).

A second explanation is that the experience of cohabitation increases the propensity to divorce. For example, Axinn and Thornton (1992) found that respondents' attitudes about the acceptability of divorce were more positive after a period of non-marital cohabitation than they were prior to a period of non-marital cohabitation. More recently, a study by Dush, Cohan and Amato (2003) found that both selection and the experience of cohabitation explanation need to be integrated to achieve a better understanding of the association between cohabitation and marriage breakdown. A third explanation relates to the length of time in the relationship (de Vaus, Qu & Weston 2005; DeMaris & Rao 1992). Generally, couples who cohabit before marriage have been in the relationship longer than those who do not cohabit. They have therefore had longer exposure to the risk of relationship dissolution, which explains in part the higher rates of dissolution observed for marriages preceded by cohabitation.

Age at marriage

In Australia, people who marry under the age of 25 have a greater risk of divorce than those who marry after age 25, irrespective of the year they married (Australian Bureau of Statistics 2000). Some argue that young age at marriage reflects a poor mate selection process because it implies that a shorter time was spent searching for a marriage partner (Becker 1981; England & Farkas 1986; Kalmijn & Poortman 2006). Others argue that these marriages face greater risk because the couples are less likely to have developed the maturity and social skills required to negotiate a long-term marital relationship and often do not have access to adequate socioeconomic and financial resources (Moore & Waite 1981; South 1995; Teti & Lamb 1989; Wolcott & Hughes 1999).

Children

Children born before marriage increase the risk of marriage breakdown (White 1990). One possible explanation is that premarital children may encourage marriage with an unsuitable partner (Waite & Lillard 1991). Alternatively, if the premarital child is not the biological child of one partner (with the exception of foster or adopted children), the child may be a destabilising force in the marriage in similar ways to step-children in remarriages (Coleman, Ganong & Fine 2000; Morgan & Rindfuss 1985).

Children born early in marriage may also increase the risk of marriage breakdown. First, an early birth may indicate a 'shotgun' wedding where a couple decides to marry because of pregnancy (Teti & Lamb 1989). This destabilises marriage because it speeds up the mate selection process and a couple may decide to marry when they otherwise would not. Second, children born early within marriage may add stress and strain in the developmental stages of the marriage (Waite & Lillard 1991). The results of studies investigating the divorce risks for marriages with early births (that is, children conceived before but born within marriage) are mixed. Becker, Landes and Michael. (1977) found premarital conception increased the risk of divorce. In contrast, Morgan and Rindfuss (1985) found that premarital conception moderately reduced the risk of marital separation, while Waite and Lillard (1991) found no significant association between premarital conception and marriage breakdown. However, given the chronology of these studies, the differences between them may also reflect changing acceptance of premarital pregnancy and birth over the last few decades.

In contrast, children born within marriage reduce the likelihood of divorce (Bracher et al. 1993; Ono 1998; Waite & Lillard 1991). It is thought that the birth of children within marriage indicates commitment to the marriage and increases its stability (Sayer & Bianchi 2000). However, the relationship between children and marriage breakdown has been found to vary depending on the ages of children. Younger children reduce the risk of marriage breakdown and older children have little or no association with the likelihood of dissolution (Ono 1998; Waite & Lillard 1991).

2.3 Attitudinal factors

Attitudes and beliefs may directly affect the risk of marriage breakdown through their influence on expectations of the conduct of marriage and family life, commitment to marriage and views on the acceptability of divorce. The two most common attitudinal measures used in the research literature on marriage breakdown are the importance of religion and gender role attitudes.

Religion

The belief systems of many religions view marriage as a sacrosanct bond and place a great deal of emphasis on remaining married, which reduces the likelihood of marriage breakdown for those who adhere to these beliefs (Waite & Lehrer 2003). Previous research has found that it is not necessarily affiliation with a particular faith but religiosity (that is, the level of engagement with religion) that is important (Call & Heaton 1997). Religiosity has a strong negative association with marriage breakdown because people who have higher levels of religiosity tend to have a stronger commitment to, and more traditional views of, marriage (Amato & Booth 1995; Greenstein 1995; Heaton & Blake 1999; Rogers & Amato 2000).

Gender role attitudes

More traditional gender role attitudes are also associated with more stable marriages. Traditional gender role attitudes uphold a gender-based division of labour and support the consequent interdependence and power differentials between spouses, where husbands are the primary breadwinners and wives are the primary homemakers and carers. Non-traditional gender role attitudes emphasise shared and negotiated divisions of labour where both spouses participate in economic activities and home-based activities with an emphasis on shared power.

There is also evidence that gender differences exist in the association between gender role attitudes and marriage breakdown. Heaton and Blake (1999) found that marriages where wives disagreed with the statement 'household tasks should be shared equally' were less likely to end in separation. Conversely, when husbands held the same view there was an increased risk of marriage breakdown. Similarly, Amato and Booth (1995) found that wives with non-traditional beliefs reported poor marital quality, but husbands with non-traditional beliefs reported good marital quality. Together this evidence suggests that men's non-traditional gender role beliefs reduce the risk of marriage breakdown, but women's non-traditional gender role attitudes increase the risk.

2.4 Socioeconomic factors

Arguably the most tested and examined structural explanation for the increase in marriage breakdown in western countries is the overall improvement in women's socioeconomic position relative to men's, particularly over the second half of the 20th century. It has been argued that wives' increased access to economic resources results in a lower level of economic dependence on their husbands, which makes it easier for them to leave an unsatisfactory marriage (Beck & Beck-Gernsheim 1995; Blau 1964; Cherlin 1992). A husband's lack of economic resources and a wife's access to economic resources or economic independence are destabilising for marriage because economic resources increase women's alternatives outside of marriage (Ono 1998; Sayer & Bianchi 2000).

Research consistently shows that men of higher socioeconomic status have more stable marriages than men of lower socioeconomic status (Corley & Woods 1991; Ono 1998; Sayer & Bianchi 2000; Tzeng & Mare 1995; Waite & Lillard 1991; White 1990). Most researchers attribute the greater stability of marriages where the husband has a higher socioeconomic position to the greater financial stability of the household and the successful fulfilment of traditional gender roles (Ono 1998). When husbands have good access to socioeconomic resources, this increases wives' benefits of staying, increases the costs of leaving and decreases the attractiveness of possible alternatives.

A large body of research has investigated the economic independence hypothesis for women using a broad range of socioeconomic measures that include employment status, hours of work, income, education, work experience and access to welfare (Hoffman & Duncan 1995; Sayer & Bianchi 2000; South 2001; Tzeng & Mare 1995). For comparability, I only included studies examining large, nationally representative population samples in the review. Overall the findings are mixed. Some studies find a significant positive association between women's economic resources and divorce (Bradbury & Norris 2005; Hoffman & Duncan 1995; Sayer & Bianchi 2000; South 2001; Tzeng & Mare 1995), others find no association (Bracher et al. 1993), and still others report a negative association, where women's higher socioeconomic position reduces the risk of divorce (Hoem 1997; Ono 1998). Evidence from Australia also indicates mixed support for the independence hypothesis. Bradbury and Norris (2005) found that couples where the wife had less than secondary education were more likely to separate. In contrast, Bracher et al. (1993) found no association between women's education and marriage breakdown.

Given the mixed support for the women's economic independence hypothesis, researchers have argued that there are scenarios where women's economic contribution to households has a stabilising rather than a disruptive effect on marriage. To illustrate, Oppenheimer (1994) has observed that men's real income has dropped over the last few decades, although in Australia this is mostly true for men at the bottom of the income distribution (Borland 1999). In addition, there is more variation in working conditions across the workforce than was once the case, and more sectors of the workforce demonstrate employment arrangements and working conditions that are not good, such as non-standard hours and casual employment (Pocock 2005). Therefore, for many marriages the economic contributions of wives may provide important resources and offer financial stability in uncertain labour markets, thereby reducing the risk of marriage breakdown. Further, as the participation of married women and mothers in paid employment increases and becomes 'normal', it is likely that the impact of women's workforce participation on marriage breakdown will be attenuated (Jalovaara 2003). There are a number of studies that offer strong empirical support for this view (for example, Ono 1998; Rogers & DeBoer 2001; Sayer & Bianchi 2000). Under these scenarios, wives' economic resources relative to their husbands' are seen as essential to understanding marriage breakdown.

2.5 The importance of gender in marriage, family life and marriage breakdown

Historically, the majority of divorce research does not account for gender. In a review article, White (1990) argued that prior to 1990 little research examined sex differences in the divorce process. Even beyond the 1990s, many divorce studies examined women only, either because relationship data were only collected from women (Bracher et al. 1993; Sweeney & Phillips 2004) or because men's reports of relationship dissolution were found to be unreliable (Bumpass, Martin & Sweet 1991). Nevertheless, over the last two decades, with the emergence of longitudinal panel studies of couples along with advances in statistical modelling, this tendency has changed. Panel data on couples enable indicators of both husbands' and wives' social characteristics to be modelled when predicting marriage breakdown. A growing body of work taking into account gender in the social determinants of marriage breakdown has emerged, particularly from the United States (see, for example, Amato & Previti 2003; Heaton & Blake 1999; Rogers & Amato 2000; Sanchez & Gager 2000; Sayer & Bianchi 2000; South 2001; Teachman 2002b; Waite & Lillard 1991), but also from Europe (see, for example, Cooke 2004; Jalovaara 2003).

The overwhelming majority of this work examines socioeconomic position of husbands and wives and finds some important gender differences. Men with a higher socioeconomic position have more stable marriages compared to men with a lower socioeconomic position (Ono 1998; Sayer & Bianchi 2000; Tzeng & Mare 1995). On the other hand, the association between women's socioeconomic position and marriage breakdown is less straightforward. For women, this association varies depending on the indicator used (Jalovaara 2003), whether or not change occurs within the marriage (Tzeng & Mare 1995) and marriage duration (South 2001). Therefore, there are gender differences in how socioeconomic characteristics impact on marriage breakdown, but the exact nature and extent of those differences is unclear.

A smaller but growing body of work investigates gender differences in other, non-socioeconomic, social predictors of marriage breakdown such as family background, religion, and relationship and fertility characteristics (Call & Heaton 1997; Heaton & Blake 1999). For example, Heaton and Blake (1999), using couple data from the National Survey of Families and Households in the United States, found that women's social background characteristics were more strongly associated with marriage breakdown than men's. Specifically, their results indicated that the wife's age at marriage, parental divorce, prior marriage and income all had a stronger association with marriage breakdown than the same characteristics for husbands.

The research literature on gender differences in the social determinants of divorce is relatively small—a much larger body of work highlights differences in men's and women's experiences of marriage and family life (Ferree 1990; Fox & Murry 2000; Thompson & Walker 1995). In general, this work examines gender differences in two main areas: psychosocial factors, and the division of paid and unpaid labour.

Research into the psychosocial aspects of divorce indicates that wives take more responsibility for the emotional and relational aspects of the marriage than do husbands (Steil 1997) and also finds gender differences in the determinants of marital quality and satisfaction (Faulkner, Davey & Davey 2005; Steil 1997; Thompson & Walker 1989). Differences between husbands' and wives' experiences and perceptions of their marital relationships have implications for marriage breakdown. Studies find that women have more complaints about their marriages and cite different reasons for their marriages ending compared to men (Amato & Previti 2003; Burns 1984; Kitson 1982; Ponzetti et al. 1992; Wolcott & Hughes 1999). For example, Amato and Previti (2003) found that women are more likely than men to report infidelity, mental or physical abuse and alcohol or drug use as reasons for divorce, and are less likely to report communication problems or to claim that they do not know what caused the divorce. Further, other research has found that women's added responsibility for relationships extends to women terminating bad marriages (Walzer & Oles 2003).

Significant gender differences also exist around the division of labour within households. Women do much more housework than men. Recent Australian research has found that women spend more than twice as much time in domestic chores as men (Baxter, Hewitt & Western 2005; Bittman et al. 2003). These trends are consistent with research from the United States (Bianchi et al. 2000) and Europe (Yodanis 2005). Further, within married couple households, the types of tasks undertaken are stratified by gender (Ferree 1991). In marriages with children, the evidence overwhelmingly suggests that even though both mothers' and fathers' time spent performing child care has increased over the last few decades, mothers continue to spend much more time in child care tasks than fathers (Baxter 2005; Craig 2006; Gaulthier, Smeeding & Furstenberg 2004; Milkie et al. 2002; Sandberg & Hofferth 2005; Sayer, Bianchi & Robinson 2004). Surprisingly little research has investigated the associations between divisions of labour and marriage breakdown, although research finds that husbands' and wives' relative paid and unpaid work is important for marriage breakdown (Cooke 2004). Other research suggests, however, that whether or not divisions of labour are associated with divorce is mediated by whether or not wives—but not husbands—perceive this division to be fair (Frisco & Williams 2003).

In sum, the research on gender, marriage and family life suggests that husbands' and wives' perceptions and experiences are different on almost every dimension of the marital relationship. Women tend to be more responsible for, and invest more in, the emotional maintenance of the relationship, care of children, and the day-to-day functioning of the household. Men, on the other hand, tend to contribute less to the running of the household but invest more in the financial stability of the household through their paid employment and continued enhancement of their economic resources. England and Farkas (1986) have observed that the main investments that men make in relationships are in resources that are useful or easily transferable outside of their current relationship, whereas women make greater relationship-specific investments. These gender differences have implications for marriage breakdown because it is women's rather than men's monitoring of relationships and satisfaction with divisions of labour that increase, or decrease, the likelihood of marriage breakdown. In light of all these factors, it is not surprising that one of the most salient gender differences in the divorce literature is that women are twice as likely as men to end their marriages (Braver, Whitely & Ng 1993; Colburn, Lin & Moore 1992; Pettit & Bloom 1984; Ponzetti et al. 1992; Sweeney 2002; Wang & Amato 2000). Nevertheless, even though many studies find that women are more likely to end their marriages than men, little research has attempted to better understand this process (for some notable exceptions, see England, Sayer & Allison 2005; Kalmijn & Poortman 2006; Pettit & Bloom 1984; Walzer & Oles 2003).

The majority of studies investigating gender differences in marriage breakdown examine how men's and women's (or husbands' and wives') characteristics are differentially associated with whether a marriage ends or remains intact. In this paper, I take an alternative approach to examining gender in the process of marriage breakdown and consider whether a marriage is ended by the husband or wife (or in some cases jointly). Bernard (1972) famously argued that within marriage there are two distinct perceptions and experiences of the relationship, 'his' and 'hers'. I argue not only is there 'his' and 'her' marriage, but there may also be 'his' and 'her' marriage breakdown.

2.6 Who initiates separation?

Over the past three decades researchers have found that initiator status (that is, which spouse initiates separation) is an important factor in developing knowledge and understanding of the processes relating to marriage breakdown. Most research examining initiator status investigates how recovery from marriage breakdown varies for initiators and non-initiators. Comparatively few studies examine differences between initiators and non-initiators prior to separation, but in common with the broader literature on marriage breakdown, there are two main approaches to investigating which spouse initiates separation. Studies investigate either the psychosocial or sociostructural determinants of initiator status, although the vast majority of studies examine psychosocial processes—in particular, people's perceptions of why their relationship ended and their feelings about their previous relationship.

In general, the body of research into the psychosocial predictors of initiator status indicates that important differences exist between initiators and non-initiators. For example, initiators, compared to non-initiators, have more positive attitudes towards divorce, and identify more alternatives to marriage (Black et al. 1991); give more, and different, reasons for why their marriages ended (Kincaid & Caldwell 1995; Pettit & Bloom 1984); report lower levels of marital quality and relationship intimacy (Vannoy 2000b); and show differences in personality traits and self-perception (Vannoy 2000a).

Some gender differences between initiators and non-initiators have been found. For example, compared to male initiators, female initiators cited more and different reasons for marriage breakdown (Pettit & Bloom 1984). In a qualitative study, Walzer and Oles (2003, p. 340) report that some women who took the initiative to end their marriages did not necessarily want to claim the status of having initiated the divorce. These women gave collective reasons for ending their marriages, such as they were acting to minimise harm to their children, or framed their decision to divorce as a last resort in response to their husbands' behaviour, such as having an affair. Men, on the other hand, often talk in initiator rhetoric even when they did not actively end their marriages (Walzer & Oles 2003, p. 342). These gender differences suggest that even though the majority of separations and divorces are initiated by wives, women often account for their actions in terms of taking care of and responsibility for the wellbeing of the family. Walzer and Oles (2003, p. 347) conclude that women in this sense are 'doing gender' by 'doing divorce', where they are taking a disproportionate responsibility for cleaning up family messes, including a bad marriage.

Only recently has research emerged that examines the associations between husbands' and wives' social characteristics and which spouse initiates separation, with studies from the United States (England, Sayer & Allison 2005; Rogers 2004) and the Netherlands (Kalmijn & Poortman 2006). This research typically uses large, nationally representative population samples. Overall, these studies provide good evidence that under certain social conditions wives rather than husbands are more likely to take the initiative to separate, while under other conditions husbands are likely to initiate separation. These studies provide consistent evidence that wives with better access to economic resources have an increased likelihood of initiating separation (England, Sayer & Allison 2005; Kalmijn & Poortman 2006), although Rogers (2004) also found an increased likelihood of husbands initiating separation when wives had better access to economic resources. The other social correlates considered by these studies are more diverse and therefore the findings are less consistent, although when husbands are younger than wives there is an increased likelihood of husbands initiating separation (England, Sayer & Allison 2005; Kalmijn & Poortman 2006). Clearly there is a need for further research into the social determinants of which spouse initiates separation.

These studies also provide good evidence that marriage breakdown is not a uniform event. In both studies, the majority of respondents—90 per cent in the Dutch study (Kalmijn & Poortman 2006) and 77 per cent in the US study (England, Sayer & Allison 2005)—reported that the separation was initiated unilaterally by either the husband or wife. Previous Australian research also finds this is the case. Wolcott and Hughes (1999) found that 79 per cent of the separations in the Australian Divorce Transitions Project were reported to have been initiated by one spouse. Moreover, given that more women than men initiate separation, it appears that the process of separation is different for husbands and wives. This paper begins to explore this possibility by examining the sociostructural predictors of marriage breakdown and which spouse initiated separation.

[ Return to Top   Return to Section ]

3. Data and methods

3.1 Data

The data used in this analysis are version 3 of Wave 1 of the Household Income and Labour Dynamics in Australia (HILDA) survey's public access confidentialised data. These data were collected in 2001 and the Wave 1, version 3 incorporates respondent information from Waves 2 and 3 where applicable. HILDA is an Australian national longitudinal panel survey, with Wave 1 comprising 7,682 households and 13,914 individuals. In Wave 1, households were selected using a multi-stage sampling approach, and a 66 per cent response rate was achieved (Watson & Wooden 2002b). Within households, data were collected from each person aged over 15 years (where available) using face-to-face interviews and self-completed questionnaires, and a 92 per cent response rate was achieved (Watson & Wooden 2002b). Overall, the HILDA sample is representative of Australian households, although there are some characteristics of individuals in the sample that differ from the general Australian population. Specifically, women are overrepresented, unmarried people are underrepresented and immigrants from non–English speaking backgrounds are underrepresented. However, the discrepancies are not large and are unlikely to compromise the overall quality of the data (Watson & Wooden 2002a).

Ideally, as indicated in Section 2, to investigate the complex issues raised by theoretical and empirical explanations for marriage breakdown, nationally representative longitudinal couple data on a wide range of sociostructural and marital characteristics for large numbers of couples before and after their marriages break down are required. There are three main reasons why this project did not use the longitudinal data from HILDA. First, this was a three-year project that began when there was only one wave of HILDA available. At best the analysis could have used three waves of HILDA; however, over the three waves there were very few marriages and divorces among the sample population. Second, because there was no prospective data collected on which spouse initiated separation until the fifth wave of data collection, I could not investigate the issue of which spouse initiated separation longitudinally, which was one of the key questions I wanted to address. In addition, there are advantages to examining a larger number of separations over a long period of time as opposed to a smaller number of very recent marriages and separations. As more waves of HILDA become available, there will be more opportunities to undertake longitudinal research into the social determinants of marriage breakdown in Australia.

3.2 The analytic sample

The analytic sample includes all respondents who had ever married (n=9,744), regardless of their marital status at the time of the survey, with some restrictions. First, people whose marriages ended in separation less than one year before the survey were excluded (n=97, <1 per cent). Previous research indicates that most marriages that reconcile tend to do so within the first 12 months of separation and including this group in the sample might have led to an overestimation of the incidence of marriage breakdown (Bumpass, Martin & Sweet 1991). Further, under the Family Law Act 1975, the only ground for divorce is irreconcilable differences evidenced by one year of separation (Stewart & Harrison 1982). Therefore all separated respondents in the sample were considered permanently separated and legally eligible to divorce.

Second, people who married in the year of the survey were excluded (n=163, 1.7 per cent), because the smallest time unit of marriage duration observed in this study is one year and this group of respondents had married less than one year prior to the survey. Third, only first marriages were considered because evidence shows that higher order marriages have an increased risk of ending and the determinants of marriage breakdown in higher order marriages tend to be different than those for first marriages (Booth & Edwards 1992; Coleman, Ganong & Fine 2000). Comparing the differences between first and higher order marriages is outside the scope of this study. Fourth, I excluded migrants who separated and/or divorced prior to living in Australia (n=154, 1.6 per cent).2 They were omitted because I was interested in the determinants of marriage breakdown in Australia, and these marriages had ended prior to the respondent/s living in Australia, presumably under different social, economic and cultural conditions.

Finally, respondents with missing marital history data (n=221, 2.3 per cent) or current marital status (n=4) were dropped from the sample. In addition, a total of 129 respondents (1.3 per cent) with missing or implausible data on some of the independent variables were dropped where the numbers were insufficient to warrant imputing or controlling for missing data. Specifically, the missing values included religion (n=13), cohabitation (n=10), father's occupation (n=2), parental divorce (n=16), children (n=17) and age at marriage (n=71).3 The final sample comprised 8,993 first marriages (4,110 men and 4,883 women).

The analyses are also restricted to only including respondents' characteristics for those who had separated or divorced in the analytic sample. The HILDA marital history data do not provide information on a respondent's spouse(s) from former marriage(s). This is a limitation for this research because the characteristics of both partners are likely to be relevant to the breakdown of marriage.

3.3 Variables

Dependent variables

Two dependent variables are examined in this paper: marriage breakdown and initiator status of marriage breakdown.

Marriage breakdown

Retrospective marriage histories were used to construct the dependent variable, marital breakdown, which is coded 0 if the respondent was still in their first marriage and 1 if their first marriage ended in separation (for at least one year) or divorce. Table 1 shows the number of people in the sample who separated or divorced. In total, 2,049 people in the sample separated or divorced from their first marriage. While most respondents (85 per cent) in the sample who separated from their first marriage went on to legally divorce, I included separation, as well as divorce, as the indicator of marriage breakdown because excluding them would have led to an underestimation of the prevalence of marriage breakdown in the sample by 15 per cent.

Table 1: Distribution of separated and divorced in analytic sample
Marital status Number %
Separated (for at least one year) 309 15
Divorced 1,740 85
Total 2,049 100

Source: HILDA Wave 1.

Initiator status

The same retrospective marriage histories were used to construct a measure of who initiated separation. In the event of separation respondents were asked: 'Whose decision was it to finally separate?' The response categories were 'Mostly mine', 'Mostly partner's', and 'Joint'. I coded these responses into three types of separation. Men who answered 'Mostly partner's' were coded 1=wife initiated and those who answered 'Mostly mine' were coded 2=husband initiated. Women who responded 'Mostly mine' were coded 1=wife initiated and women who answered 'Mostly partner's' were coded 2=husband initiated. If either men or women answered 'Joint' that response was coded 3=jointly initiated. In this final category, both spouses were considered to play a major role in the final decision to separate from the marriage. People still in their first marriage were coded 0=still married.

Two methodological issues relating to the measurement of initiator status have been identified in the literature. First, defining initiator status is difficult and there are several different aspects of the breakdown of a marriage that can potentially be construed as 'initiation'. Prior studies use a variety of measures for initiator status, but a study by Braver et al. (1993), which examined three different measures of initiator status, found that they are not necessarily interchangeable. For example, the spouse who first suggested divorce is not necessarily the same spouse who filed the legal papers for divorce. This measure identifies the partner who made the final decision to separate from the marriage, and that person is not always the same partner who filed for divorce, physically left the relationship or first raised the issue of divorce.

A second methodological issue is the potential for systematic bias in the reporting of initiator status. Research finds an ego-enhancing bias in reports of who initiated separation, with respondents more likely to report they, not their former spouses, initiated the separation (Amato & Previti 2003). Further, research investigating the level of consistency in the reporting of initiator status between former spouses shows that there is close, but not perfect, agreement between reports; Braver, Whitely and Ng (1993) found that 70 per cent of former spouses agreed on who initiated the marriage breakdown, and Sweeney (2002) found agreement in 80 per cent of cases.

In the current sample, further analysis was undertaken to better understand the nature of the bias in the sample by investigating the effects of time since separation on the differences between women's and men's responses as to who initiated the separation. This strategy was based on the assumption that recall bias would be greater the longer the amount of time since separation. Results from this analysis showed no significant differences between women's and men's reports of initiator status if they had separated up to two years before the survey, but there were significant differences between those separated longer than two years prior to survey. To adjust for this bias (and reduce the impact of reporting bias on the estimates) a dummy control, coded 1=separated less than two years, and 2=not separated and separated two or more years, was included in the models examining initiator status. It is worth noting here that the inclusion of this measure into the models resulted in changes for some variable coefficients at the second or third decimal point level, but there was no change in the substantive interpretation of the results.

Independent variables

Table 2 describes the independent variables. I included a range of temporal, family background, relationship and fertility, attitudinal and socioeconomic measures in the analyses. The temporal measure is birth cohort of respondent. It consists of 11 five-year cohorts, with the oldest cohort born prior to 1925 and the youngest born after 1971. Birth cohort is included in the model as a series of dummy variables, with the middle cohort (1946–1950) chosen as the reference category to accommodate the curvilinear relationship between divorce and birth cohort.

There are three family background measures. The first is the respondent's ethnicity, coded as 1=Australian born, 2=overseas born—ESB (English-speaking background), and 3=overseas born—NESB (non–English speaking background), with Australian born as the reference group. I also included a measure of mother's and father's occupational status, each with five categories: 1=managers and administrators, 2=professionals, 3=white collar, 4=blue collar and 5=never worked. The reference category is managers and administrators, and a dummy for missing values for mother's and father's occupations is included. The final measure of family background is a dummy for parental divorce (1=yes).

There are four measures of relationship and fertility history. I included a dummy for cohabitation prior to marriage (1=yes), and a dummy for whether or not the respondent had a child prior to marriage (1=yes). Age at marriage is included as a continuous variable. The final measure is a time-varying indicator for whether or not any children were born within the marriage, coded 0 before any children were born within the marriage and coded 1 from the year the first child was born within the marriage. A dummy was also included for missing data on the child measures.

The attitudinal measures are religiosity and gender roles. Religiosity is indicated by 'importance of religion' to the respondent, measured by a scale ranging from 0=not important to 10=very important. The measure for gender role attitudes indicates agreement with the statement that: 'It is much better for everyone involved if the man earns the money and the woman takes care of home and children'. The responses were on a scale ranging from 1=strongly disagree to 7=strongly agree; hence a higher score indicates more traditional gender role attitudes.4 There are some limitations with using both of these indicators as they are measured at the date of survey, after the point of separation for all separated respondents.5 This is problematic because it is unclear if the associations found are due to respondents' attitudes changing after separation or divorce, rather than reflecting their attitudes prior to separation or divorce. Any problems this poses for interpretation of the results are addressed in Section 4.2.

Table 2: Description of covariates, by sex
Covariate

Women (n=4,883)

Men (n=4,110)

Mean SD(a) Mean SD(a)
Temporal and family background
Birth cohort        
<1925 0.07   0.06  

1926–1930

0.06   0.06  

1931–1935

0.06   0.06  

1936–1940

0.07   0.08  

1941–1945

0.08   0.09  

1946–1950

0.10   0.11  

1951–1955

0.11   0.12  

1956–1960

0.13   0.13  

1961–1965

0.13   0.13  

1966–1970

0.11   0.10  

>1971

0.09   0.06  
Ethnicity        

Australian born

0.74   0.71  

Overseas born—ESB

0.10   0.12  

Overseas born—NESB

0.16   0.17  
Father's occupation        

Manager/administrator

0.20   0.21  

Professional

0.22   0.21  

White collar

0.11   0.09  

Blue collar

0.44   0.44  

Never worked

0.03   0.00  

Missing

0.04   0.04  
Mother's occupation        

Manager/administrator

0.03   0.03  

Professional

0.15   0.15  

White collar

0.26   0.25  

Blue collar

0.24   0.22  

Never worked

0.28   0.29  

Missing

0.04   0.06  
Parents ever divorced (1=yes) 0.16   0.13  
Relationship and fertility
Cohabited prior to marriage (1=yes) 0.29   0.30  
Age at marriage 22.94 4.5 25.44 5.0
Premarital birth (1=yes) 0.07   0.06  
Early birth 0.06   0.07  
First child born in marriage (1=yes)(b) 0.65   0.67  
Missing values for children measures 0.09   0.06  
Attitudinal
Religiosity 5.52 3.5 4.49 3.6
Gender role attitudes 4.06 2.1 3.71 2.0
Missing 0.08   0.08  
Socioeconomic
Highest level of education        

Bachelor degree or higher

0.20   0.20  

Diploma

0.08   0.10  

Trade or certificate

0.22   0.35  

Year 12 or less

0.47   0.34  

Missing

0.03   0.02  

(a) Standard deviations are only reported for continuous measures. (b) Indicates that measure is time varying. Note: SD=standard deviation; ESB=English-speaking background; NESB=non–English speaking background. Source: Primary data analysis of HILDA Wave 1, version 3 data.

Education is the measure of socioeconomic position. I used education because previous research has shown that it is a relatively stable indicator of socioeconomic position that tends to be established early in adulthood and changes very little after marriage (Tzeng & Mare 1995). Other measures of socioeconomic position such as occupation and income are more volatile during marriage and retrospective data are not available. Level of education consists of four categories: 1=bachelor degree or higher, 2=diploma, 3=trade/certificate and 4=Year 12 or less. Bachelor degree or higher is the reference group and a dummy variable for missing values is also included.

3.4 Analytic approach

The first dependent variable, marriage breakdown, is an event that is time dependent. I therefore used a discrete time survival analysis modeling approach, which takes into account this time dependency (Singer & Willett 2003). To do this, I constructed a marriage-year dataset, where each respondent contributed one person-year to the dataset for every year they were married. The dependent variable was coded 0 in years that a respondent was married and 1 in the year they separated. Respondents who remained married until surveyed were coded 0 only on the dependent variable and treated as censored. If a respondent's marriage ended in an interval due to the death of a spouse, the case was treated as censored. In other words, one-year marriage intervals were included for that respondent until the year they were widowed. Logistic regression was then used to predict the likelihood that a marriage would end given that it did not end in the previous year. Likelihood ratio and Wald tests were used to establish whether each independent variable was associated with marital breakdown. I adjusted for duration dependence in the model using a fifth order polynomial. Even though there is a broadly curvilinear trend (suggesting a quadratic baseline hazard) there are a number of 'ups' and 'downs' in the hazard rate at certain marriage durations that suggest that the trend is not strictly linear or quadratic.

It should be noted that many of the respondents were currently married. For the majority of these married respondents, data on both spouses were included in the sample. This should not pose too much of an issue as most analyses were run separately for men and women, so husbands and wives were treated as separate observations and the observations should therefore be random to the extent that the households were random. While most of the models were run separately for men and women to establish whether there are gender differences, for some additional analysis (that is, the gender interactions models testing whether the gender differences are statistically significant) both men and women were included in the models. In these circumstances I also adjusted for clustering within households because 'still married' men and women were from the same household. The 'cluster' command in Stata specifies that the observations are independent across groups (that is, respondents from different households) but not necessarily within groups (that is, respondents from the same household) (StataCorp 2003). This command estimates a robust standard error and variance–covariance matrix for the explanatory variables but does not affect the coefficients. Therefore, applying this command to the models adjusts the significance level accounting for covariance between spouses at the household level but does not adjust the coefficients.

[ Return to Top   Return to Section ]

4. Marriage breakdown in Australia

Whose marriages breakdown? In this section I use the existing international and national literature on the social correlates of marriage breakdown as a framework to investigate the patterns of associations between social characteristics and marriage breakdown in the sample of first marriages from HILDA. I also take account of possible gender differences in the patterns of association in two ways. First, I conduct all analyses separately for men and women, and second, I formally test whether the observed gender differences are significant by pooling men's and women's samples and including interaction terms between gender with all covariates.

4.1 Survival and hazard estimates of marriage breakdown

Survival estimates

Figure 1 plots the survival estimates, which is the proportion of marriages that were still intact (surviving) by marriage duration, from 0 to 68 years, for the sample. This figure shows the overall proportion of men and women who had not experienced marriage breakdown since entry into the risk set. The overall survival rates of first marriages in the analytic sample were 68.1 per cent for women and 68.7 per cent for men; hence marriage survival rates are essentially identical for men and women in the sample. The survival probabilities drop steadily within the first 25 years or so of marriage for women and men, after which the survival function flattens out. After 35 or 40 years the proportion staying married is relatively constant and does not drop below 68 per cent for either men or women, although this final figure varies slightly for men and women. It is interesting to note that this estimate from the HILDA data accords closely with official estimates that 32 per cent of current marriages will end in divorce (Australian Bureau of Statistics 2000).

Figure 1: Survival estimates of marriage breakdown for analytic sample, by sex

Figure 1: Survival estimates of marriage breakdown for analytic sample, by sex

Source: HILDA Wave 1, Version 3.

Hazard estimates

Figure 2 shows the hazard rates of marriage breakdown for men and women. The hazard rate indicates the proportion of marriages that ended due to separation for each time interval, given that the respondent had entered into that time interval. Overall, these results demonstrate a broadly curvilinear association between length of marriage and marital breakdown in the sample. The hazard of separation (marriage breakdown) increases within the first five years of marriage and then declines at a decreasing rate in subsequent years. The hazard of separation after 40 years of marriage is virtually non-existent. In any given year, the probability of separation is very low for both women and men, but as the survival estimates indicate, the cumulative incidence of marriage breakdown over time is much greater. These results also indicate that overall the men and women in the sample have very similar patterns of marriage duration and probability of marital breakdown.

Figure 2: Hazard rate of marriage breakdown by duration of marriage in analytic sample, by sex

Figure 2: Hazard rate of marriage breakdown by duration of marriage in analytic sample, by sex

Source: HILDA Wave 1, Version 3.

4.2 Social factors associated with marriage breakdown

Next, I examine the associations between social characteristics and whether or not the marriage ended in separation. Table 3 presents the odds ratio for all predictors except the polynomial term for duration dependence, and missing values. A value greater than one indicates a multiplicative increase in the odds of marriage breakdown as the covariate increases, while a value less than one indicates a multiplicative decrease in the odds of marriage breakdown as the covariate increases (Box-Steffensmeir & Jones 2004).

Table 3: Discrete time event history model predicting the risk of marriage breakdown as a function of various social characteristics(a) for men and women
Social characteristic Women Odds ratio Men Odds ratio
Temporal and family background
Birth cohort(b), (c)    

<1925

0.30*** 0.29***

1926–1930

0.37*** 0.35***

1931–1935

0.48*** 0.62**

1936–1940

0.59*** 0.73*

1941–1945

0.75* 0.83

1946–1950 (reference group)

1.00 1.00

1951–1955

1.01 1.10

1956–1960

1.15 1.09

1961–1965

1.18 1.37

1966–1970

1.20 1.20

>1971

1.24 0.97
Ethnicity(c), (d)    

Australian born

1.00 1.00

Overseas born—ESB

1.18 0.77*

Overseas born—NESB

0.80* 0.92
Father's occupation    

Manager/administrator

1.00 1.00

Professional

1.06 1.33*

White collar

1.03 1.09

Blue collar

1.08 1.22

Never worked

1.63 0.83
Mother's occupation    

Manager/administrator

1.00 1.00

Professional

1.52* 1.29

White collar

1.50 1.17

Blue collar

1.28 1.15

Never worked

1.20 1.13
Parents ever divorced(b), (c) 1.51*** 1.24*
Relationship and fertility
Cohabited prior to marriage(b), (c) 1.33*** 1.44***
Age at marriage(b), (c), (d) 0.93*** 0.96***
Premarital birth(b), (c) 2.35*** 1.78***
Early birth(b), (c) 1.32* 1.67***
First child born in marriage(b), (c), (e) 0.37*** 0.41***
Attitudinal
Religiosity(b), (c) 0.96*** 0.96***
Gender role attitudes 1.03* 1.04
Socioeconomic
Highest level of education(b), (d)    

Bachelor degree or higher

1.00 1.00

Diploma

0.89 1.61***

Trade or certificate

0.97 1.39**

Year 12 or less

0.79* 1.39**
Number of respondents 4,883 4,110
Marriage-years 101,232 85,081
Number of events 1,147 902

(a) Models also include polynomial terms for duration dependence and controls for missing values for some measures. (b) Likelihood ratio tests indicate that the inclusion of this variable in the model resulted in a significant improvement in the overall fit of the model at p<0.05 for men (Appendix A). (c) Likelihood ratio tests indicate that the inclusion of this variable in the model resulted in a significant improvement in the overall fit of the model at p<0.05 for women (Appendix A). Note: ESB=English-speaking background; NESB=non–English speaking background. * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Source: Primary data analysis of HILDA Wave 1, version 3 data.

Temporal and family background factors

Consistent with previous Australian research (Bracher et al. 1993), men and women in birth cohorts older than the reference cohort had a significantly decreased likelihood of marriage breakdown. This supports the argument that older cohorts have been exposed to different historical and contextual influences that result in a lower propensity to divorce compared to younger cohorts (Hackstaff 1999).

This analysis shows that men from other English-speaking countries have a significantly lower risk of marriage breakdown than Australian-born men. This result is not consistent with previous Australian research which reports that English-speaking migrants have an increased risk of divorce (Bracher et al. 1993; de Vaus 1997, 2004). This association is not significant for women, although the gender interactions suggest that migrant women from English-speaking countries have a significantly greater risk of marriage breakdown than Australian-born. Again, this contradicts previous findings which suggest that the likelihood of divorce in these groups should be similar for men and women (de Vaus 2004). In contrast, the results suggest that women from non–English speaking countries have a comparatively lower risk of marriage breakdown than Australian-born women. This is consistent with previous Australian research. For example, Bracher et al. (1993) found that immigrant women from non–English speaking backgrounds, particularly those of southern European origin, were less likely to experience marriage breakdown than Australian-born women.

Overall, parent's occupation was not a strong predictor of marriage breakdown. Men whose fathers had professional occupations had a significantly increased risk of marriage breakdown compared to men whose fathers were managers or administrators. This is inconsistent with a US study by Corley and Woods (1991) which found that men whose fathers had high occupational status had a longer duration between marriage and divorce. Women whose mothers had professional occupations had a significantly increased risk of marriage breakdown than women whose mothers were managers or administrators. Research on mother's occupation and the likelihood of divorce is scant, but the result here is consistent with the expectation that women whose mothers have high occupational status may be more likely to experience marriage breakdown due to the non-traditional gender role these mothers model.

Consistent with prior research from Australia (Bracher et al. 1993), the United States (McLanahan & Bumpass 1988; Mueller & Pope 1977; Teachman 2004), Canada (Hall & Zhao 1995) and Britain (Kiernan & Cherlin 1999), the results here suggest that parental divorce increases the risk of marriage breakdown for both men and women. There are two main explanations for this association. First, children of divorced parents, compared to children whose parents remained married, exhibit social and life course characteristics that increase the risk of divorce, such as young age at marriage, premarital childbirth and cohabitation prior to marriage (Kiernan & Cherlin 1999; McLanahan & Bumpass 1988; Mueller & Pope 1977; Teachman 2002a, 2004). Second, children of divorced parents are not as well equipped to negotiate long-term relationships compared to children whose parents remained married (Amato 1996). For example, in an Australian study Burns and Dunlop (2000) reported that children of divorced parents had more behavioural problems compared with children of intact families, which in turn had a negative impact on the quality of their intimate relationships 10 years later.

Relationship and fertility factors

Cohabitation before marriage increases the risk of marriage breakdown for both men and women, a finding that agrees with prior Australian (Sarantakos 1994), US (Bumpass, Martin & Sweet 1991) and Canadian (Hall & Zhao 1995) research. Both men and women who marry at younger ages have an increased likelihood of marriage breakdown than those who marry older. This finding is consistent with the maturity hypothesis that those who marry at younger ages tend to have less maturity and life experience to successfully negotiate a marital relationship. In addition, the gender interactions indicate that this association is significantly stronger for women. This may be because women who marry young are more likely to initiate separation than men. If younger age at marriage indicates a poorer-quality match (Becker 1981) and women are more sensitive to relationship quality (Steil 1997; Thompson & Walker 1989), then women may be more likely to initiate separation in marriages characterised by young age at marriage. Over the last few decades in Australia there has been a strong trend towards marrying later: in 1982 the median age at first marriage was 25 for men and 22 for women; by 2003 this had risen to 29 for men and 27 for women (Australian Bureau of Statistics 2005). This suggests that age at marriage may not continue to be as important for future generations, although it is likely that this increase in age at marriage is due in part to the increasing rates of cohabitation prior to marriage in recent marriage cohorts.

Consistent with previous research, the results here indicate that both premarital birth and early birth increase the risk of marriage breakdown for men and women. It is also worth noting here that early birth has a much greater effect on men's odds of separation than women's, and the gender interactions model indicates that this gender difference nearly reaches statistical significance at p=0.072. This may be because some men feel 'trapped' into marrying because their partners were pregnant and therefore may be more likely to initiate separation in these circumstances than women. The finding that marriages are much less likely to end when the first child is born within marriage is consistent with prior studies (Cherlin 1977; Thornton 1977).

Attitudinal factors

Consistent with previous work on the association between religion and marriage breakdown (Call & Heaton 1997; Lillard, Brien & Waite 1995), the results here indicate that religiosity reduces the likelihood of marriage breakdown for both men and women. Call and Heaton (1997) found that high levels of religious commitment among husbands and wives reduced the risk of marriage breakdown, and when spouses reported extremely different levels of religiosity the marriage was less stable. This suggests that the combination of spouses' religiosity, not just one spouse's religiosity, is also important for marriage stability. Further, the causal direction of this association is somewhat uncertain as the measure of religion here is taken at the time of the survey, after separation or divorce, and it is possible the experience of marriage breakdown may influence religiosity. This issue is worthy of further research.

Even though gender role attitudes reach statistical significance for women, gender role attitudes are not important predictors of marriage breakdown in the current analysis as they did not improve the overall fit of the model. This contradicts previous research, using a single-item measure similar to that used here, which found that marriages where wives have traditional gender role attitudes and husbands have non-traditional gender role attitudes are significantly less likely to end in separation or divorce (Heaton & Blake 1999). This finding in the current study may be because the measure included is the respondent's gender role attitude at the time of survey, which is post-separation and divorce, whereas attitudes are likely to change over the life course, including in response to major life events such as separation and divorce. The influence of gender role attitudes on marriage breakdown is best captured with longitudinal data on couples before and after separation. Further, other measures included in the models, such as religiosity and cohabitation, to some extent, act as proxy measures for gender roles, which may be attenuating the overall effects of gender roles.

Socioeconomic factors

The results here suggest that men with higher levels of education have a decreased risk of separation compared to those with low levels of education, but highly educated women have an increased risk of separation. This finding for women contradicts other studies that find better educated women tend to be less likely to separate or divorce (Jalovaara 2003; South 2001; Tzeng & Mare 1995). While other socioeconomic measures may provide different results, education is the best available measure of socioeconomic status because there were no retrospective data available on occupation, income and employment histories. A more complete test of the association between socioeconomic factors and marriage breakdown would examine a broader range of measures and compare partners' socioeconomic positions in determining the risk of marital breakdown. This modelling could not be pursued here because of data limitations.

4.3 Summary

These results confirm the importance, and significance, of social characteristics in understanding marriage breakdown. The patterns of association between temporal, life course, attitudinal and economic factors identified in the international literature are consistent with these Australian data. In general, the patterning of association between social background factors and marriage breakdown was similar for men and women, but the findings also suggest some gender differences in the association of age at marriage, ethnic background and education. These gender differences may be due to differences in the decisions of men and women to initiate separation when these characteristics are present in marriage. While it is not possible to examine the relative influence of couple characteristics on marriage breakdown with the available data, gender differences in the process by which the marriage ended can be investigated by examining which spouse initiated separation. In the next section, I further explore the gendered nature of marriage breakdown by using social characteristics to predict whether the marriage ended because the wife, husband or both spouses took the initiative to end the relationship.

[ Return to Top   Return to Section ]

5. Who initiates separation?

When marriages end, what is the likelihood that wives rather than husbands took the initiative and made the decision to separate? In this section, I first describe the number of men and women (husbands and wives) who ended their marriages. Using a similar set of social characteristics to the previous analysis, I then investigate whether the social characteristics identified in the literature predict whether men or women are more likely to end their relationship. The analysis in this section takes into account gender in the process of marriage breakdown by investigating whether the wife, husband or both spouses decided to initiate separation.

5.1 Reports of which spouse initiated separation

Table 4 shows the distribution of initiator status (that is, which spouse decided to initiate the separation) for men and women in the analytic sample. In total, 902 men and 1,147 women in the sample separated or divorced from their first marriage. Of these, men reported wife-initiated separation in 35 per cent of cases, husband-initiated separation in 27 per cent of cases and jointly initiated separation in 37 per cent of cases. Women reported wife-initiated separation in 58 per cent of cases, husband-initiated separation in 18 per cent of cases, and jointly initiated separation in 24 per cent of cases. It is worth noting here that, in total, only 30 per cent of separations were reported to have been jointly initiated. In the other 70 per cent of cases, initiator status is assigned to either the wife or the husband. Of these separations, wives initiated in 69 per cent of cases.

Table 4: Initiator status of separation for separated and divorced men and women
 

Women

Men

  Total
Number % Number %   Number %
Wife initiated 665 58 320 35   985 48
Husband initiated 201 18 245 27   446 22
Jointly initiated 281 24 337 37   618 30
Total 1,147 100 902 100   2,049 100

Source: HILDA Wave 1 (2001).

The overall figures from this study differ from other comparable studies. In the Netherlands, Kalmijn and Poortman (2006) found that 61 per cent of separations were initiated by wives, 29 per cent were initiated by husbands, and only 10 per cent were jointly initiated. The figures here are more similar to the US study by England, Sayer and Allison (2005), which found that wives initiated in 53 per cent of cases and husbands in 24 per cent, with 23 per cent of cases defined as 'other', which included jointly initiated and cases where former husbands and wives disagreed about which spouse initiated separation. Despite these overall differences, the one consistent finding across all studies and countries is that wives are around two times more likely than husbands to initiate separation.

5.2 Hazard estimates of which spouse initiated separation

Figure 3 shows the hazard rate of marriage breakdown for men and women in the sample. While overall rates of marriage breakdown are similar for men and women, there are some subtle differences in the timing of separations initiated by wives and husbands according to men's and women's reports. For men, during the first few years of marriage, the probability of reporting a wife-initiated separation is lower than a husband or jointly initiated separation. After that the probability of a wife-initiated separation is slightly higher than that for a husband-initiated separation. For women, on the other hand, the probability of reporting a wife-initiated separation is higher than reporting a husband-initiated separation until around 30 years of marriage. The patterning for jointly initiated separations is similar for men and women, with an increasing probability in the first few years of marriage followed by a steep decline and plateauing after about 30 years of marriage.

Figure 3: Hazard rates of marriage breakdown and men's and women's reports' of who initiated separation

Figure 3: Hazard rates of marriage breakdown and men's and women's reports' of who initiated separation

Note: The raw conditional probabilities have been smoothed with a locally weighted regression of the probability of marriage breakdown on marriage duration (StataCorp 2003). Source: HILDA Wave 1. Source: HILDA Wave 1.

5.3 Social correlates of which spouse initiates separation

Next, I examine whether there are differences in the social characteristics associated with men's and women's decisions to separate. For this analysis, I excluded those measures that did not significantly improve the fit of the model in the previous section, namely father's and mother's occupations and gender role attitudes. The results are presented separately for men and women.

Table 5 shows the associations between women's social characteristics and their reports of who initiated separation relative to staying married. Overall, few women's characteristics are associated with their reports of husbands initiating separation. At first glance, the patterning of associations for women's characteristics appears to be different for separations initiated by wives compared to those initiated by husbands or jointly.

Overall, regardless of who initiated separation, women in older cohorts are less likely to experience marriage breakdown than younger cohorts. The results for ethnic background indicate that, compared to Australian-born women, immigrant women from English-speaking countries are at significantly greater risk of a wife-initiated separation, rather than staying married. In contrast, women from non–English speaking countries have a reduced risk of separations initiated by husbands or wives. Women's ethnicity is not significantly associated with jointly initiated separation. Women whose parents divorced report an increased risk of wife and jointly initiated separation, but there is no evidence that a wife's parental divorce is associated with an increased risk of separation initiated by her husband.

Table 5: Discrete time event history models with competing risks of women's reports of who initiated separation relative to still married by social characteristics
Social characteristic Wife initiated Odds ratio Husband initiated Odds ratio Jointly initiated Odds ratio
Temporal and family background
Birth cohort      

<1925

0.34*** 0.34* 0.16**

1926–1930

0.36*** 0.16** 0.63

1931–1935

0.49*** 0.48 0.46*

1936–1940

0.52** 0.84 0.66

1941–1945

0.88 0.94 0.42**

1946–1950

1.00 1.00 1.00

1951–1955

0.87 1.24 1.27

1956–1960

1.14 1.43 1.07

1961–1965

1.28 1.05 1.00

1966–1970

1.15 0.82 1.45*

>1971

1.28 0.33 1.40
Ethnicity      

Australian

1.00 1.00 1.00

Overseas born—ESB

1.29*(a) 1.03 1.19

Overseas born—NESB

0.75* 0.62* 0.88
Parents ever divorced(b) 1.64*** 1.42 1.49*
Relationship and fertility
Cohabited prior to marriage(b) 1.28* 1.28 1.36
Age at marriage 0.91***(a) 0.98 0.95**
Premarital birth 2.29*** 1.36 2.51***(a)
Early birth 1.27 0.94 2.07***
First child born in marriage(c) 0.37*** 0.51* 0.33***
Attitudinal
Religiosity(d) 0.94*** 0.98(a) 0.96*
Socioeconomic
Highest level of education      

Bachelor degree or higher

1.00 1.00 1.00

Diploma

1.00 0.98 0.44**(a)

Trade or certificate

0.96(a) 0.97 0.81

Year 12 or less

0.84(a) 0.77 0.51***(a)
Number of respondents   4,883  
Marriage-years   101,232  
Number of separations 665 201 281
Goodness of fit LR Chi2 (df)   1,112.97 (90)  

(a) Results of pooled gender interactions model (Appendix C) indicate that this is statistically different (p<0.05) from the same characteristic for men (Table 6). (b) 0=no, 1=yes. (c) Indicates that measure is time varying. (d) Scale ranging from 0=not important to 10=very important. Notes: Table does not report dummies for missing data or the controls for duration dependence (fifth order polynomial) and reporting bias. ESB=English-speaking background; NESB=non–English speaking background. LR Chi2 (df )=likelihood ratio with a chi-square distribution and degrees of freedom (in brackets). * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Source: Primary data analysis of HILDA Wave 1, version 3 data.

Cohabiting prior to marriage increases the risk of wife-initiated separation. In relation to age at marriage, marrying later significantly decreases a woman's risk of separation initiated either by her or jointly, but is not associated with whether her husband initiates separation. Women who had a premarital birth have significantly increased odds of reporting wife or jointly initiated separation relative to staying married, but premarital birth is not associated with reports of husbands initiating separation. Early birth significantly increases the risk of reporting a jointly initiated separation, but not separations initiated by husbands or wives. When the first child is born within the marriage, there is a reduced likelihood of all types of separation, although the magnitude of this association is smaller for women's reports of husbands initiating separation than of wife or jointly initiated separations.

Religiosity is significantly associated with separation initiated by wives and jointly initiated separation, with the risk of wife and jointly initiated separation, compared to staying married, declining with women's increased religiosity. There is no association between women's religiosity and separations initiated by husbands. Finally, women's education is only significantly associated with jointly initiated separation. Women with a diploma, or less than Year 12 education, have a significantly lower risk of jointly initiated separation than women with a bachelor degree.

The results for men are presented in Table 6. The association between birth cohort and separation is similar to that for women: irrespective of their reports of who initiated separation, men born in older cohorts have a lower risk of marriage breakdown than the reference cohort. Men's ethnic background is not associated with which spouse initiated separation. Yet the results of the gender interactions model (see Appendix C) show that the lower risk of wife-initiated separation found for men from English-speaking countries is significantly different from the large increased risk of women from English-speaking countries reporting a wife-initiated separation (Table 5). Men whose parents divorced, compared to men whose parents stayed married, report an increased risk of wives initiating separation, but parental divorce is not linked to a husband or jointly initiated separation.

Men who cohabited before marriage have a greater risk of all types of separation. Older age at marriage significantly reduces the risk of husband and jointly initiated separations, but not separations initiated by wives. The results of the gender interactions model suggest that the small, non-significant association between age at marriage and men's reports of a wife-initiated separation, and the large negative association between age at marriage and women's reports of a wife-initiated separation are significantly different. Premarital and early births also increase the risk of separation. A premarital birth increases the risk of men reporting a husband-initiated separation, but not a wife or jointly initiated separation. Early birth increases the risk of wife and jointly initiated separation, but not husband-initiated separation. Having a first child born in marriage reduces the risk of separation generally.

Men's religiosity is significantly and negatively associated with husband and jointly initiated separations, but is not associated with men's reports of separations initiated by wives. Given that religiosity is measured after separation or divorce, this may reflect a change in religiosity after divorce.

Finally, men with diploma and trade/certificate levels of education are more likely than men with a bachelor degree or higher qualification to report that their wives initiated separation. The results for diploma level education also suggest that men with a diploma qualification are more likely to initiate separation than men with a bachelor degree or higher. Together the results suggest that relative to men with a bachelor degree or higher qualification that those with a diploma level of education have an increased risk of separation initiated by either themselves or their wives. The gender interactions suggest that the difference between lower-educated men's increased risk of wife and jointly initiated separation and lower-educated women's decreased chances of wife and jointly initiated separation are statistically significant.

Table 6: Discrete time event history models with competing risks of men's reports of who initiated separation relative to still married by social characteristics
Social characteristic Wife initiated Odds ratio Husband initiated Odds ratio Jointly initiated Odds ratio
Temporal and family background
Birth cohort      

<1925

0.28** 0.39* 0.25***

1926–1930

0.34** 0.48* 0.25***

1931–1935

0.73 0.77 0.40**

1936–1940

0.62 0.75 0.84

1941–1945

0.79 0.94 0.77

1946–1950

1.00 1.00 1.00

1951–1955

1.14 1.11 0.95

1956–1960

1.02 1.07 1.15

1961–1965

1.37 1.10 1.37

1966–1970

1.01 0.96 1.11

>1971

1.10 0.81 1.18
Ethnicity      

Australian born

1.00 1.00 1.00

Overseas born—ESB

0.94(a) 0.89 0.80

Overseas born—NESB

1.02 1.03 0.91
Parents ever divorced(b) 1.42* 1.10 1.33
Relationship and fertility
Cohabited prior to marriage(b) 1.43* 1.64** 1.37*
Age at marriage 0.98(a) 0.95* 0.95**
Premarital birth 1.54 2.36*** 1.23(a)
Early birth 1.54* 1.26 1.96***
First child born in marriage(c) 0.47*** 0.50** 0.31***
Attitudinal
Religiosity(d) 0.98 0.92***(a) 0.95**
Socioeconomic
Highest level of education:      

Bachelor degree or higher

1.00 1.00 1.00

Diploma

1.60*(a) 1.93** 1.38(a)

Trade or certificate

1.68**(a) 1.11 1.22

Year 12 or less

1.44 1.31 1.28(a)
Number of respondents   4,110  
Marriage-years   85,081  
Number of separations 320 245 337
Goodness of fit LR Chi2 (df )   1,125.56 (90)  

(a) Results of pooled gender interactions model (Appendix C) indicate that this is statistically different at p<0.05 from the same characteristic for women (Table 5). (b) 0=no, 1=yes. (c) Indicates that measure is time varying. (d) Scale ranging from 0=not important to 10=very important. Notes: ESB=English-speaking background; NESB=non–English speaking background. LR Chi2 (df )=likelihood ratio with a chi-square distribution and degrees of freedom (in brackets). * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Table does not include dummies for missing values or the controls for duration dependence (fifth order polynomial) and reporting bias. Source: Primary data analysis of HILDA Wave 1, version 3 data.

Next, I investigate whether the differences between separations initiated by wives and husbands are statistically significant. In some respects, this set of results provides a better test of the differences between the likelihood of wife-initiated separation compared to husband-initiated separation. This is because the significance of the effects for the results presented in Tables 5 and 6 depend, in part, on the number of events observed for each type of separation. Therefore, the effects of social characteristics on a wife-initiated separation are more likely to be significant than for a husband or jointly initiated separation, particularly for women. This next set of results tests for differences between the equations and provides important information about differences between separations initiated by wives compared to those initiated by husbands or jointly. I re-estimate the same model, but the contrast category on the dependent variable is wife-initiated separation, indicating the conditional probability of a husband-initiated separation versus a wife-initiated separation or a jointly initiated separation versus a wife-initiated separation. For brevity, only statistically significant results for both sexes are presented in Table 7.

The results for women are presented on the left-hand side of Table 7. Only two characteristics significantly differentiate between women's reports of separations initiated by them relative to separations initiated by their husbands or jointly: age at marriage and education. The findings for age at marriage suggest that older age at marriage reduces women's likelihood of reporting separation initiated by them, but not by their husbands. In addition, the gender interactions model indicates that the difference between women's reports of an increased risk of wife compared to husband-initiated separation, and men's reports of a lower risk of wife compared to husband-initiated separation, is statistically significant. This provides some additional support for the finding reported in Table 5 that women's older age at marriage significantly reduced the risk of separations initiated by wives, but had no significant association for separations initiated by husbands. Women's education differentiates between wife and jointly initiated separations, but not husband-initiated separations. Compared to women with a bachelor degree or higher, women with diplomas or less than high school education have significantly lower odds of reporting a jointly initiated separation than a wife-initiated separation.

Table 7: Competing risks model, contrasting wife-initiated separation by selected social characteristics for women and men
Social characteristic

Women's reports

Men's reports

Husband initiated Odds ratio Jointly initiated Odds ratio Husband initiated Odds ratio Jointly initiated Odds ratio
Attitudinal
Religiosity(a) 1.03 1.01 0.93**(b) 0.96
Relationship and fertility
Age at marriage 1.06*(c) 1.04 0.97 0.97
Socioeconomic
Highest level of education        

Bachelor degree or higher

1.00 1.00 1.00 1.00

Diploma

0.98 0.44** 1.21 0.86

Trade or certificate

1.01 0.85 0.66 0.72

Year 12 or less

0.92 0.61* 0.91 0.89
Number of respondents

4,883

4,110

Marriage-years

101,232

85,081

Number of separations 200 279 244 334
Goodness of fit LR Chi2 (df )

1,112.97 (90)

1,125.56 (90)

(a) Scale ranging from 0=not important to 10=very important. (b) Results of pooled gender interactions model (Appendix C) indicate that this is statistically different from women's reports of a husband compared to a wife-initiated separation. (c) Pooled gender interactions model (Appendix C) indicates that this is statistically different from men's reports of a husband compared to a wife-initiated separation. Note: The LR Chi2 (df )=likelihood ratio with a chi-square distribution and degrees of freedom. * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Source: Primary data analysis of HILDA Wave 1, version 3 data.

The results for men's reports are presented in the right-hand side of Table 7. The only social factor that distinguishes between men's reports of wives or husbands initiating separation is religiosity. Men who are more religious have a reduced risk of reporting that they initiated separation compared to wives than less religious men. The gender interactions further indicate that the significant negative association between men's religiosity and husband-initiated separation is statistically different from the small, non-significant association between women's religiosity and husband-initiated separation. Overall, even though the results of the first analysis indicated differences between wife-initiated and husband or jointly initiated separation for several social characteristics, the results of this analysis suggest that most were not statistically significant.

5.4 Summary

The main gender difference observed in the results is that women's characteristics are strongly associated with their reports of wife-initiated separation and not husband-initiated separation, but this trend is not mirrored in men's reports. Men's characteristics are relatively evenly associated with their reports of both husband and wife-initiated separation. This suggests that separations initiated by wives are associated with both men's and women's characteristics, but separations initiated by husbands are primarily associated with men's characteristics.

This finding is consistent with qualitative research which found that in some circumstances wives initiate separation because their husbands are unhappy, or because they do not want their children to be exposed to a bad marriage, not because they themselves are necessarily unhappy with the marriage (Hackstaff 1999; Walzer & Oles 2003). It is also worth noting that part of the failure to find significant predictors of wife rather than husband-initiated separations for women may be because there are only a small number of husband-initiated separations reported by women in the sample.

The findings also suggest, however, that some social characteristics vary the likelihood of husbands initiating separation compared to wives. Men's ethnic background is not significantly associated with any specific type of separation. In contrast, women born overseas in English-speaking countries have an increased likelihood of reporting a wife-initiated separation, but not a husband-initiated separation. This implies that immigrant women from English-speaking countries are more likely to take the initiative to separate. In contrast, women from non–English speaking backgrounds report a significantly lower risk of separations initiated by both husbands and wives.

Both women and men whose parents divorced have an increased likelihood of reporting wife-initiated separation, but parental divorce is not significantly associated with husband-initiated separation. This might suggest that wives may be more likely to initiate separation in difficult marriages, regardless of whether the instability is associated with parental divorce on the husband's or wife's side.

According to women's reports, older age at marriage reduces the risk of wife and jointly initiated separation, while according to men's reports, older age at marriage reduces the risk of husband and jointly initiated separation. Similar to the findings in relation to religiosity, these results imply that it is the respondent's age and not their partner's which increases the risk of initiating separation. In addition, women who married younger are significantly more likely to report a wife-initiated separation than men who married younger. This is consistent with the 'maturity' explanation proposed earlier in Section 2.2: if young age at marriage is associated with poor mate selection and wives' marital satisfaction is tied more strongly to the intimate and emotional qualities of the relationship than husbands', women who marry at younger ages have an increased likelihood of initiating separation (Moore & Waite 1981; Teti & Lamb 1989; Wolcott & Hughes 1999).

Men's religiosity decreased the likelihood of reporting a husband-initiated separation; likewise, women's religiosity reduced the likelihood of reporting a wife-initiated separation. These results do not suggest differences between husband and wife-initiated separations, but rather imply that the importance of religion shapes an individual's, as opposed to their spouse's, decisions about continuing their marriage. However, the results of the gender interactions model (Appendix C) indicate that the association between religiosity and the chance of reporting a husband-initiated separation rather than a jointly initiated separation is weaker for men than women. Therefore, the magnitude of the association between religiosity and marriage breakdown appears to be much greater for husbands compared to wives. It should be noted that the causal direction of these results is unclear, as the measure was taken at the time of survey, not prior to separation or divorce. It is possible that the importance of religion to a person may change in the event of separation and divorce, and these results may reflect changes after separation and divorce.

Using the highest level of education as an indicator of socioeconomic position, the results indicate that women with lower levels of education have a reduced likelihood of jointly initiated separations, but not separations initiated by husbands or wives. This suggests that more educated women are more likely to report a jointly initiated separation than less educated women. Men with lower education levels are more likely to report that their wives initiated separation, while men with diploma-level education have an increased risk of husband-initiated separation.

[ Return to Top   Return to Section ]

6. Conclusions

Overall, the findings of this paper can be grouped into three main themes: first, sociostructural factors are important for understanding why some marriages break down and others remain intact; second, wives are more likely to initiate marital separation than husbands; and third, while some gender differences in the social correlates of which spouse initiated separation are present, in general men and women tend to end their marriages under similar circumstances. The main gender difference is that husbands' characteristics are associated with wives ending marriages, but wives' characteristics are not related to husbands' decisions to end their marriages.

6.1 Social factors are important for understanding marriage breakdown

Both men and women are less likely to experience marriage breakdown if their normative and cultural social characteristics (such as religiosity, birth cohort, ethnic background and cohabitation before marriage) reflect more traditional family organisation and greater commitment to the institution of marriage. Similarly, social characteristics that imply a poorer mate selection process such as young age at marriage or early birth, or a decrease in the ability of one or both spouses to negotiate the relationship, such as parental divorce, tend to increase the risk of marriage breakdown for men and women. Marital children reduce the likelihood of marriage breakdown for both men and women. However, highest level of education operates differently for men and women. Men with higher levels of education have a lower risk of marriage breakdown than lower-educated men, but higher levels of education for women increase the risk of marriage breakdown compared to lower-educated women.

Further, the results suggest that some characteristics influence marriage breakdown through more than one mechanism. For example, ethnic background may cause disruption because of differences in normative and cultural expectations of marriage, or because migration places additional stresses and strains on a relationship that increase the likelihood of disruption. This suggests that sociostructural predictors of divorce operate in both direct and indirect ways to influence marriage outcomes. Untangling some of these complexities is an important direction for future research.

Consistent with previous studies that examined differences in the association between husbands' and wives' social characteristics and marriage breakdown (Amato & Previti 2003; Heaton & Blake 1999; Rogers & Amato 2000; Sanchez & Gager 2000). I found some gender differences. For women, young age at marriage and having migrated from an English-speaking country are associated with an increased risk of marriage breakdown compared to men. For men, having lower levels of education is associated with a greater risk of marriage breakdown compared to more educated men. However, for women this association is the opposite: less educated women have a lower risk of marriage breakdown. Research from the United States and Europe (Cooke 2004; Sayer & Bianchi 2000; South 2001) suggests that it is the relative characteristics of couples, not just the individual characteristics that each partner brings to the relationship, that contribute to marriage breakdown. Unfortunately, testing this further was outside the scope of this project. However, as future waves of HILDA become available with observations on people entering marriage and separating from marriage, this will be an important direction for future research.

6.2 Women are more likely to initiate separation than men

Overall, I found that women are more likely to initiate marital separation than men. At the same time, there are important differences between men's and women's reports of who initiated separation. Women are significantly more likely to report wife-initiated separation than men and, conversely, men are significantly more likely to report husband or jointly initiated separation than women. The baseline hazards also indicate gender differences: according to women's reports, wives are much more likely to initiate separation early in marriage than husbands. According to men's reports, however, the three types of separation are very similar across marriage duration.

This ego-enhanced reporting bias is similar to that found in other studies (Kalmijn & Poortman 2006; Wang & Amato 2000), but assuming that the truth is somewhere in the middle, wives are still more likely to initiate separation than husbands. In total, taking into account both men's and women's reports, around 70 per cent of separations are initiated by either the husband or wife. Of these separations, 69 per cent were wife initiated and 31 per cent were husband initiated. These overall figures are similar to those of the other two comparable studies. In the United States, England, Sayer and Allison (2005) found that 77 per cent of separations were initiated unilaterally and of these 69 per cent were initiated by wives. In the Netherlands, Kalmijn and Poortman (2006) report that 90 per cent of separations were initiated unilaterally with 68 per cent of these initiated by wives. In the Australian Divorce Transition Project, 79 per cent of separations were initiated unilaterally, with 75 per cent of these initiated by wives (Wolcott and Hughes 1999). The consistent finding across these studies, and countries, is that around two-thirds of unilateral separations are initiated by wives.

6.3 Gender and initiator status

One of the key issues to be addressed in this paper was whether men and women initiate separation under different circumstances. More particularly, are some conditions associated with wives ending marriages, while other conditions are associated with husbands initiating separation? The answer is both yes and no. For some social characteristics, the patterning of association suggests that wives are more likely to initiate separation than husbands, while other characteristics are associated with husbands initiating separation. Specifically, women are more likely to initiate separation than comparable men when they migrate from English-speaking countries, when they marry young and when they have higher levels of education. Men with higher levels of attachment to religion are less likely to initiate separation than comparable women. Although there are exceptions, the general pattern is that where there are significant gender differences they point to an increased likelihood of women initiating separation.

Overall, however, there are few significant gender differences and when differences are found they are often in the magnitude of the effect rather than in the direction of the effect. Essentially similar characteristics are associated with men and women ending their relationships. The main gender difference observed is that, according to women's reports, very few women's social characteristics are associated with husbands initiating separation. This patterning of results is not evident for men, whose characteristics are more evenly associated with their reports of both husbands and wives initiating separation. What this suggests is that wives seem to initiate separation on the basis of their husbands' characteristics, whereas men tend to initiate separation mainly on the basis of their own characteristics. This is consistent with family research investigating gender differences in marriage and family life, which has found that women's perceptions of relationship quality and satisfaction are tied more to intimacy and emotional qualities of the relationship (Faulkner, Davey & Davey 2005). Women also take more responsibility for, and spend more time, investing in the maintenance of the relationship and family life (England & Farkas 1986; Steil 1997). These two mechanisms may, at least in part, explain these findings.

Qualitative research investigating gender differences in the divorce process indicates that some wives initiate separation because their partners are unhappy, or because they do not want their children to be exposed to a bad marriage, rather than because they themselves are necessarily unhappy with the marriage, whereas husbands tend to initiate separation when they are unhappy (Hackstaff 1999; Walzer & Oles 2003). Therefore, women are more likely to initiate separation because they are more sensitive to and aware of relationship problems. There is good evidence that this is the case. Research has found that women have more complaints about their marriages than men (Amato & Previti 2003; Burns 1984; Ponzetti et al. 1992).

In addition, because women typically invest more in, and take greater responsibility for the maintenance of, marital relationships than men they are more likely to feel that they are not getting satisfactory returns from marriage, and it is possible that this contributes to their increased likelihood of initiating separation (Beck-Gernsheim 2002; Giddens 1992). For example, in some marriages husbands indirectly end the marriage by behaving in ways—such as openly having an affair or spending more time with their friends than their families—that force their wives to end the marriage (Hetherington & Kelly 2002; Hopper 1993; Walzer & Oles 2003). Therefore, women are more likely to initiate for multiple reasons to do with their perceptions of marital satisfaction and quality, their investments in relationships and their greater responsibility for marriage and family life. Men on the other hand tend to initiate when they are unhappy with the relationship but tend not to monitor their wives' level of happiness with the marriage (Walzer & Oles 2003). Hence, men are more likely than women to report that they 'don't know' why their marriage ended (Amato & Previti 2003; Wolcott & Hughes 1999).

6.4 Policy discussion

From this and other research (Bracher et al. 1993; White 1990), it is clear that social contexts and individual social characteristics are important for understanding why some marriages remain intact and other marriages break down. It is also clear that women are more likely to initiate separation than men (Braver, Whitely & Ng 1993; England, Sayer & Allison 2005; Kalmijn & Poortman 2006; Wolcott & Hughes 1999). Some have argued that divorce is too easy, and that the forces of individualism and feminism are undermining the foundations of family life (see, for example, House of Representatives Standing Committee on Legal and Constitutional Affairs 1998, pp. 55–64). Implicit in these arguments is that a return to earlier, more traditional models of marriage and family life entailing a husband–father breadwinner and a wife–mother homemaker will solve the 'divorce problem'. Even though marriage rates were also low in the first half of the 20th century, it is likely that divorce rates in Australia were kept 'artificially' low through cultural norms and divorce and labour market policies that reinforced women's dependence on marriage, and men. These forces would have kept many unhappy, unproductive marriages intact. It is not plausible to return to policies that reinforced women's dependence on men and marriage for several reasons.

First, Australian research indicates that men's wages have declined over the last few decades, particularly for men in lower wage groups, and greater proportions of employed men are working casually and part-time. In contrast, women's wages have increased and more women are working full-time (Birrell, Rapson & Monash University Centre for Population and Urban Research 1998). Many families would find it very difficult to survive financially without the earnings of women, so policies that restrict women's workforce participation would not be good for the sustainability of family life (Oppenheimer 1994). Further, while the evidence on whether or not women's workforce participation increases the risk of marriage breakdown is far from clear, it is clear that women's paid work can help stabilise family life (Ono 1998; Oppenheimer 1994, 1997). Also, the normative and cultural climate surrounding women's employment has changed, and continues to change. The role of employee is now an accepted and expected part of women's life course. Recent US research shows that men consider women's socioeconomic prospects when looking for a marriage partner (Sweeney & Cancian 2004).

Second, there have been cultural changes in expectations of marriage and family life, with a more egalitarian relationship ideal emerging in which both spouses in a marriage share in paid and unpaid work and constantly negotiate arrangements that suit both their needs (Beck & Beck-Gernsheim 1995; Beck-Gernsheim 2002; Giddens 1992), although research suggests that most marital relationships fall far short of this ideal (Gross & Simmons 2002). Finally, restricting women's access to alternatives to marriage has trapped some women and children in physically and emotionally abusive situations (Nock, Wright & Sanchez 1999).

So, given that it is impossible to go back, how do we go forward? In Australia, at the time of this study, there is policy pressure in two main directions that, it is believed, will help preserve marriage and family life and reduce marriage breakdown: the first is to implement labour market and workplace reforms that take the pressure off marriage and family life; and the second is to amend marriage and divorce legislation and policy. An example of the first is the recent discussion paper titled Striking the balance: women, men, work and family (Human Rights and Equal Opportunity Commission 2005). Acknowledging that dual-income households have become the norm and that women's workforce participation is essential to household financial stability and the Australian economy, the report recommends a range of work and labour market-related reforms designed to help couples balance work, marriage and family life. Some of the recommendations include improving both men's and women's access to flexible work arrangements and parental leave entitlements, increasing affordable day care places, and providing tax incentives for dual part-time employed households.

There has also been pressure to change marriage legislation and divorce policy in ways designed to reduce the rate of divorce. Following the lead of some US states including Louisiana, Arizona and Arkansas, the possibility of introducing alternative, more strict, marriage contracts, such as covenant marriage, in Australia was raised some years ago (House of Representatives Standing Committee on Legal and Constitutional Affairs 1998, p. 202). Covenant marriage contracts differ from regular marriage contracts in several ways: couples are required to undertake premarital counselling and participate in marriage counselling before being allowed to divorce; and divorce cannot be obtained until two years of separation have passed, unless extenuating circumstances can be proved, such as abuse or adultery (Brinig 1998; Nock, Wright & Sanchez 1999; Sanchez et al. 2002). In the three US states, covenant marriage has not replaced current marriage laws, but rather is an alternative for those wanting to make a stronger statement about their marriage (Rosier & Feld 2000). Research on the US laws indicates that covenant marriage has not been as popular as proponents had expected (Rosier & Feld 2000) and even though it was reported at the time that up to 20 states considered introducing covenant marriage legislation, only three US states introduced the new marriage laws (Hawkins et al. 2002). Since 1998, there have not been any further suggestions to change the marriage laws in Australia.

Over the last decade or so, the primary policy direction for divorce prevention in Australia, and other developed western countries such as the United States, has been to focus on encouraging couples to use premarital education and marriage counselling services (Halford & Simons 2005; Markman & Halford 2005). This direction is being driven by good evidence that skills-based premarital education can increase marriage quality and reduce the likelihood of divorce in couples from a broad range of backgrounds (Stanley et al. 2006) and that high-quality marriage counselling can help couples work through relationship difficulties (Markman & Halford 2005). The problem has been in getting couples to participate, particularly couples with the highest risk of divorce. Australian research by Halford et al. (2006) indicates that couples who are the most at risk of divorce are also the least likely to seek help when their marriage is troubled. It is important to note, however, that the results of this study are based on a survey of people divorcing in Queensland, Australia, which only achieved a 7 per cent response rate. Therefore, while the study provides good indications of possible associations, the results cannot necessarily be extrapolated to the broader population. Further investigation into the links between social risk factors, premarital education, marriage counselling and marriage breakdown in Australia is needed. While the research in this paper is useful for identifying at-risk groups in the community, no links between social characteristics and marriage education and counselling can be made. Beyond identifying at-risk groups, marketing and information campaigns need to be targeted to these groups to encourage participation in premarital education, and marriage counselling during marriage.

Despite the positive policy directions for marriage education and counselling, there are some contradictions in modern Australian society that are likely to limit the success of these attempts to reduce marriage breakdown. On the one hand, the Australian Government is seeking to strengthen marriage and families and encourage couples to stay together. On the other hand, current labour market policies encourage competition and individualism in the pursuit of workplace success. Work intensification and polarised working hours are now a feature of Australian family life in many couple households (Western, Baxter & Chesters 2007). These two forces are inherently contradictory. Encouraging and supporting couples to deal with marital issues at a dyadic level, without changing other structural features of modern society, including, but not limited to, work and labour market policies, may limit the impact of policy initiatives aimed at preventing marriage breakdown. A key theme driving this paper has been to illustrate how strongly marriage outcomes are linked to broader social structures and that gender differences in the opportunities and constraints of those structures are also important for understanding why some marriages end while others remain intact. Taken together, the findings of this project indicate that marriage outcomes will not change without complementary changes in the broader social and structural fabric of Australian society.

[ Return to Top   Return to Section ]

Appendix A: Likelihood ratio tests comparing model in Table 3 with models excluding each of the covariates

Table A1: Likelihood ratio tests comparing model in Table 3 with models excluding each of the covariates
  Men LR—test (df ) Women LR—test (df )
Main model (LR Chi2 (df )) 606.64 (41) 992.33 (41)
Excluding birth cohort 81.01 (10)*** 99.02 (10)***
Excluding father's occupation 9.28 (5) 7.64 (5)
Excluding mother's occupation 2.66 (5) 10.10 (5)
Excluding ethnicity 5.51 (2) 8.84 (2)*
Excluding parental divorce 4.87 (1)* 27.60 (1)***
Excluding cohabitation prior to marriage 21.24 (1)*** 12.66 (1)***
Excluding age at marriage 14.73 (1)*** 57.93 (1)***
Excluding premarital birth 17.40 (1)*** 46.39 (1)***
Excluding early birth 19.06 (1)*** 6.35 (1)*
Excluding first child born in marriage 44.91 (1)*** 78.92 (1)***
Excluding religiosity 17.44 (1)*** 22.40 (1)***
Excluding gender role attitudes 3.94 (2) 4.01 (2)
Excluding education 18.73 (4)*** 9.10 (4)

Note: The LR Chi2 (df )=likelihood ratio with a chi-square distribution and degrees of freedom. * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001.

Source: Primary data analysis of HILDA Wave 1, version 3 data.

[ Return to Top   Return to Section ]

Appendix B: Gender interactions for discrete time event history model presented in Table 3

Table B1: Gender interactions for discrete time event history model presented in Table 3
Main effects β se
Female 1.44* 0.51
Normative and cultural    
Birth cohort    

<1925

–1.23*** 0.23

1926–1930

–1.05*** 0.20

1931–1935

–0.47** 0.17

1936–1940

–0.32* 0.14

1941–1945

–0.17 0.13

1946–1950 (reference group)

  —   —

1951–1955

0.08 0.12

1956–1960

0.06 0.13

1961–1965

0.27* 0.13

1966–1970

0.13 0.17

>1971

–0.08 0.27
Ethnicity    

Australian born

  —   —

Overseas born—ESB

–0.25* 0.11

Overseas born—NESB

–0.08 0.10
Religiosity –0.04*** 0.01

Gender role attitudes

0.04 0.02

Cohabited prior to marriage

0.34*** 0.09
Pyschosocial disruption    
Father's occupation    

Manager/administrator

  —   —

Professional

0.28* 0.12

White collar

0.08 0.14

Blue collar

0.19 0.10

Never worked

–0.20 0.10
Mother's occupation    

Manager/administrator

  —   —

Professional

0.26 0.25

White collar

0.15 0.24

Blue collar

0.15β 0.24

Never worked

0.12 0.24
Parents ever divorced (1=yes) 0.22* 0.10
Age at marriage –0.04*** 0.01
Premarital childbirth 0.59*** 0.14
Early birth 0.54*** 0.11
Social barriers    
Child born in marriage(a) –0.91*** 0.12
Highest level of education    

Bachelor degree or higher

  —   —

Diploma

0.48*** 0.13

Trade or certificate

0.33** 0.11

Year 12 or less

0.33** 0.11
Gender interactions    
Normative and cultural    
Birth cohort    

<1925 x female

0.06 0.30

1926–1930 x female

–0.07 0.28

1931–1935 x female

–0.26 0.24

1936–1940 x female

–0.20 0.21

1941–1945 x female

–0.11 0.18

1951–1955 x female

0.07 0.17

1956–1960 x female

–0.10 0.17

1961–1965 x female

–0.08 0.18

1966–1970 x female

0.09 0.21

>1971

0.32 0.32
Ethnicity    

Overseas born—ESB x female

0.41** 0.14

Overseas born—NESB x female

–0.15 0.14
Religiosity x female 0.00 0.01
Gender role attitudes x female –0.00 0.03
Cohabited prior to marriage x female –0.05 0.12
Pyschosocial disruption    
Father's occupation    

Professional x female

–0.22 0.15

White collar x female

–0.05 0.19

Blue collar x female

–0.11 0.14

Never worked x female

0.71 1.13
Mother's occupation    

Professional x female

0.17 0.33

White collar x female

0.25 0.32

Blue collar x female

0.11 0.32

Never worked x female

0.06 0.31
Parents ever divorced x female 0.20 0.12
Age at marriage x female –0.03* 0.01
Premarital childbirth x female 0.24 0.18
Early birth x female –0.29 0.16
Social barriers    
Child born in marriage(a) x female –0.08 0.16
Highest level of education    

Diploma x female

–0.59** 0.18

Trade or certificate x female

–0.37* 0.15

Year 12 or less x female

–0.56*** 0.15
Number of respondents 8,883  
Marriage-years 186,313  
Number of events 2,047  

(a) Indicates that measure is time varying. Note: Analysis also includes polynomial for duration dependence and controls for missing values on some measures. ESB=English-speaking background; NESB=non–English speaking background. β=coefficient; se=standard error of coefficient. * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Source: Primary data analysis of HILDA Wave 1, version 3 data.

[ Return to Top   Return to Section ]

Appendix C: Social characteristics interacted with gender by reports of who initiated separation (supplementing tables 5, 6 and 7)

Table C1: Social characteristics interacted with gender by reports of who initiated separation
 

Baseline category equals still married

  Baseline category equals wife initiated
 Wife initiated Husband initiated Jointly initiated   Husband initiated Jointly initiated
Main effects β se β se β se   β se β se
Female  2.67*** 0.56 –1.68 1.06 –0.17 0.84   –5.02 1.31 –3.52 1.13
Birth cohort                      

<1925

–1.28** 0.39 –0.94* 0.42 –1.37** 0.39   0.35 0.57 –0.08 0.55

1925–1930

–1.09** 0.33 –0.74* 0.37 –1.37*** 0.36   0.36 0.50 –0.28 0.49

1931–1935

0.31 0.27 –0.26 0.31 –0.92** 0.31   0.05 0.41 –0.61 0.41

1936–1940

–0.47 0.26 –0.29 0.29 –0.16 0.23   0.19 0.39 0.27 0.34

1941–1945

–0.24 0.22 –0.06 0.25 –0.27 0.21   0.18 0.34 –0.03 0.31

1946–1950 (reference group)

                     

1951–1955

0.13 0.20 0.10 0.23 –0.05 0.21   –0.03 0.31 –0.19 0.28

1956–1960

0.02 0.21 0.06 0.25 0.14 0.21   0.04 0.33 0.11 0.30

1961–1965

0.32 0.22 0.09 0.26 0.32 0.22   –0.23 0.35 –0.01 0.31

1966–1970

0.01 0.28 –0.04 0.32 0.10 0.28   –0.05 0.43 0.09 0.40

>1971

0.10 0.50 –0.21 0.56 0.16 0.42   –0.31 0.74 0.07 0.65
Ethnicity                      

Australian

 —  —  —  —  —  —    —  —  —  —

Overseas born—ESB

–0.22 0.18 –0.11 0.20 –0.22 0.18   0.11 0.27 0.00 0.26

Overseas born—NESB

–0.13 0.17 0.03 0.20 –0.10 0.17   0.16 0.26 0.03 0.24
Parents ever divorced 0.34* 0.15 0.09 0.19 0.28 0.15   –0.24 0.24 –0.05 0.21
Cohabited prior to marriage 0.36* 0.15 0.49** 0.18 0.32* 0.16   0.14 0.23 –0.04 0.22
Age at marriage –0.02 0.02 –0.05** 0.02 –0.05** 0.02   –0.03 0.03 –0.03 0.02
Premarital birth 0.43 0.23 0.86*** 0.23 0.20 0.26   0.43 0.33 –0.23 0.35
Early birth 0.43* 0.19 0.23 0.24 0.67*** 0.18   –0.20 0.30 0.25 0.26
First child born in marriage –0.76* 0.22 –0.69** 0.26 –1.18*** 0.19   0.07 0.33 –0.42 0.28
Religiosity –0.02 0.02 –0.08*** 0.02 –0.05** 0.02   –0.06 0.03 –0.04 0.02
Highest level of education:                      

Bachelor degree or higher

—   —  —  —  — —     —  —  — — 

Diploma

0.47* 0.24 0.66** 0.24 0.32 0.22   0.19 0.34 –0.15 0.32

Trade or certificate

0.52** 0.18 0.11 0.21 0.20 0.17   –0.41 0.28 –0.32 0.25

Year 12 or less

0.36 0.19 0.27 0.21 0.24 0.17   –0.09 0.28 –0.12 0.26
Gender interactions                      
Birth cohort                      

<1925 x female

0.21 0.47 –0.13 0.61 –0.45 0.68   –0.34 0.78 –0.65 0.82

1926–1930 x female

0.06 0.42 –1.12 0.69 0.91 0.49   –1.19 0.81 0.85 0.64

1931–1935 x female

–0.40 0.35 –0.47 0.49 0.15 0.46   –0.08 0.60 0.54 0.58

1936–1940 x female

–0.18 0.33 0.11 0.42 –0.24 0.37   0.28 0.54 –0.07 0.49

1941–1945 x female

0.11 0.28 –0.00 0.38 –0.61 0.38   –0.12 0.47 –0.72 0.47

1951–1955 x female

–0.28 0.25 0.11 0.35 0.30 0.30   0.39 0.43 0.57 0.39

1956–1960 x female

0.11 0.26 0.30 0.36 –0.07 0.31   0.18 0.45 –0.18 0.41

1961–1965 x female

–0.07 0.27 –0.05 0.41 –0.32 0.34   0.02 0.49 –0.25 0.44

1966–1970 x female

0.13 0.33 –0.16 0.51 0.27 0.39   –0.29 0.61 0.14 0.51

>1971 x female

0.15 0.54 –0.90 0.95 0.18 0.54   –1.05 1.09 0.03 0.76
Ethnicity                      

Overseas born—ESB x female

0.48* 0.22 0.15 0.30 0.39 0.26   –0.33 0.37 –0.09 0.34

Overseas born—NESB x female

–0.16 0.22 –0.50 0.31 –0.03 0.25   –0.34 0.38 0.14 0.34
Parents ever divorced x female 0.16 0.18 0.09 0.19 0.11 0.21   0.10 0.32 –0.05 0.28
Cohabited prior to marriage x female –0.11 0.19 0.49 0.18 –0.00 0.23   –0.14 0.33 –0.11 0.30
Age at marriage x female –0.07 0.02 –0.05 0.02 –0.01 0.02   0.10 0.04 0.06 0.03
Children                      

Premarital birth x female

0.40 0.28 –0.55 0.40 0.72 0.34   –0.95 0.48 0.32 0.44

Early birth x female

–0.18 0.25 –0.29 0.39 0.05 0.27   –0.10 0.46 0.24 0.37

First child born in marriage x female

–0.23 0.26 0.02 0.39 0.08 0.28   0.25 0.45 0.31 0.37
Religiosity x female –0.03 0.02 –0.08 0.02 0.01 0.03   0.09 0.04 0.04 0.03
Highest level of education  —  —  —  —  —  —    —  —  —  —
Diploma x female –0.47 0.29 –0.65 0.24 –1.14 0.35   –0.21 0.47 –0.67 0.45
Trade or certificate x female –0.56* 0.22 0.11 0.21 –0.41 0.24   0.42 0.38 0.16 0.33
Year 12 or less x female –0.54* 0.22 0.27 0.21 –0.92 0.25   0.004 0.38 –0.38 0.34

Note: ESB=English-speaking background; NESB=non–English speaking background. β=coefficient; se=standard error of coefficient. * significant at p<0.05; ** significant at p<0.01; *** significant at p<0.001. Source: Primary data analysis of HILDA Wave 1, version 3 data.

[ Return to Top   Return to Section ]

Endnotes

1. This figure was derived from estimated costs incurred by the Department of Social Security, the Family Court, Legal Aid, the Child Support Scheme and the Sole Parent Tax Rebate in the financial year 1994–95 due to relationship breakdown.

2. To investigate the impact of migrants who had separated prior to migration on the results, I re-estimated the model presented in Table 3. I found that including the migrant group who had separated or divorced before coming to Australia positively increased the magnitude of the coefficients for all ethnic groups; there were also some changes in the results of other covariates as the characteristics of people in those groups also changed. This is because including this group adds to the number of separated/divorced respondents in the migrant categories. I have excluded this group because they had separated (or divorced) prior to living in Australia and therefore the characteristics of their first marriage do not relate to their experiences in Australia.

3. This group of respondents also included those who had an implausible value for age at marriage which, for the purposes of this study, was considered to be marriage under the age of 16 years.

4. I acknowledge that there is one limitation of using a single-item measure for gender role attitudes, namely that there are lower levels of reliability with using a single item. HILDA Wave 1 does have a series of 14 questions relating to work and family issues. I conducted a factor analysis of these items and found that five factors emerged; however, the alphas were low for each factor with the highest alpha at 0.67. I therefore decided to include a single-item measure in the analysis. This is comparable with other studies that have investigated the associations between gender role attitudes and marriage breakdown (see, for example, Heaton & Blake 1999).

5. Before deciding to leave these measures in, I examined the impact of their inclusion and exclusion on the other model covariates. While some coefficients for the other covariate changed, the overall substantive findings remained the same. I therefore decided to leave these measures in the models.

[ Return to Top   Return to Section ]

References

Australian Bureau of Statistics 2000, 'Special article: lifetime marriage formation and marriage dissolution patterns in Australia', Marriages and divorces, cat. no. 3310.0, Australian Bureau of Statistics, Canberra, pp. 84–91.

——2005, Marriages, Australia, cat. no. 3306.0.55.001, [electronic product], Australian Bureau of Statistics, Canberra.

Amato, PR 1996, 'Explaining the intergenerational transmission of divorce', Journal of Marriage and the Family, vol. 58, pp. 628–40.

——2000, 'The consequences of divorce for adults and children', Journal of Marriage and the Family, vol. 62, pp. 1269–87.

Amato, PR & Booth, A 1995, 'Changes in gender role attitudes and perceived marital quality', American Sociological Review, vol. 60, pp. 58–66.

Amato, PR & Previti, D 2003, 'People's reasons for divorcing: gender, social class, the life course, and adjustment', Journal of Family Issues, vol. 24, pp. 602–26.

Axinn, WG & Thornton, A 1992, 'The relationship between cohabitation and divorce: selectivity or causal influence?' Demography, vol. 29, pp. 357–74.

Baxter, J 2003, 'Families and households', R Hassan, I McAllister & S Dowrick (eds), The Cambridge handbook of social sciences in Australia, Cambridge University Press, pp. 462–79.

——2005, 'To marry or not to marry: marital status and the household division of labor', Journal of Family Issues, vol. 26, pp. 300–21.

Baxter, J, Hewitt, B & Western, M 2005, 'Post-familial families and the domestic division of labour: a view from Australia', Journal of Comparative Family Studies, vol. 36, pp. 583–600.

Beck, U & Beck-Gernsheim, E 1995, The normal chaos of love, Polity Press, Cambridge.

Becker, GS 1981, A treatise on the family, Harvard University Press, Cambridge, Massachussetts.

Becker, GS, Landes, EM & Michael, RT 1977, 'An economic analysis of marital instability', The Journal of Political Economy, vol. 85, pp. 1141–88.

Beck-Gernsheim, E 2002, Reinventing the family: in search of new lifestyles, Polity Press, Cambridge.

Bernard, J 1972, The future of marriage, Yale University Press, New Haven, Connecticut.

Berrington, A & Diamond, I 2000, 'Marriage or cohabitation: a competing risks analysis of first-partnership formation among the 1958 British birth cohort', Journal of the Royal Statistical Society: Series A, vol. 163, pp. 127–51.

Bianchi, SM, Milkie, M, Sayer, LC & Robinson, JP 2000, 'Is anyone doing the housework? Trends in the gender division of household labour', Social Forces, vol. 79, pp. 191–228.

Birrell, R, Rapson, V & Monash University Centre for Population and Urban Research 1998, A not so perfect match: the growing male/female divide 1986–1996, Monash University Centre for Population and Urban Research, Clayton, Victoria.

Bittman, M, England, P, Folbre, N, Sayer, LC & Matheson, G 2003, 'When does gender trump money? Bargaining and time in household work', American Journal of Sociology, vol. 109, pp. 186–214.

Black, LE, Eastwood, MM, Sprenkle, DH & Smith, E 1991, 'An exploratory analysis of the construct of leavers versus left as it relates to Levinger's social exchange theory of attractions, barriers, and alternative attractions', Journal of Divorce and Remarriage, vol. 15, pp. 127–39.

Blau, PM 1964, Exchange and power in social life, John Wiley & Sons, New York.

Booth, A & Edwards, JN 1992, 'Starting over: why remarriages are more unstable', Journal of Family Issues, vol. 13, pp. 179–94.

Borland, J 1999, 'Earnings inequality in Australia: changes, causes and consequences', The Economic Record, vol. 75, pp. 177–202.

Box-Steffensmeir, JM & Jones, BS 2004, Event history modelling: a guide for social scientists, Cambridge University Press.

Bracher, M, Santow, G, Morgan, SP & Trussell, J 1993, 'Marriage dissolution in Australia: models and explanations', Population Studies, vol. 47, pp. 403–25.

Bradbury, B & Norris, K 2005, 'Income and separation', Journal of Sociology, vol. 41, pp. 425–46.

Braver, SL, Whitely, M & Ng, C 1993, 'Who divorced whom? Methodological and theoretical issues', Journal of Divorce and Remarriage, vol. 20, pp. 1–19.

Brinig, MF 1998, 'Economics, law, and covenant marriage', Gender Issues, vol. 16, pp. 4–33.

Bumpass, LL, Martin, TC & Sweet, JA 1991, 'The impact of family background and early marital factors on marital disruption', Journal of Family Issues, vol. 12, pp. 22–42.

Burns, A 1981, 'Divorce and the children', Australian Journal of Sex, Marriage and Family, vol. 2, pp. 17–26.

——1984, 'Perceived causes of marriage breakdown and conditions of life', Journal of Marriage and the Family, vol. 46, pp. 551–62.

Burns, A & Dunlop, R 2000, 'Parental divorce, personal characteristics and early adult intimate relationships: a longitudinal Australian study', Journal of Divorce and Remarriage, vol. 33, pp. 91–109.

Call, VRA & Heaton, TB 1997, 'Religious influence on marital stability', Journal for the Scientific Study of Religion, vol. 36, pp. 382–92.

Carmichael, GA, Webster, A & McDonald, PF 1996, Divorce Australian style: a demographic analysis, Research School of Social Sciences, Australian National University, Canberra.

Cherlin, A 1977, 'The effect of children on marital dissolution', Demography, vol. 14, pp. 265–72.

——1978, 'Remarriage as an incomplete institution', American Journal of Sociology, vol. 84, pp. 634–50.

Cherlin, AJ 1992, Marriage, divorce, remarriage, revised and enlarged, Harvard University Press, Cambridge, Massachusetts.

Colburn, K, Lin, PL & Moore, MC 1992, 'Gender and the divorce experience', Journal of Divorce and Remarriage, vol. 17, pp. 87–108.

Coleman, M, Ganong, L & Fine, M 2000, 'Reinvestigating remarriage: another decade of progress', Journal of Marriage and the Family, vol. 62, pp. 1288–307.

Cooke, LP 2004, 'The gendered division of labor and family outcomes in Germany', Journal of Marriage and the Family, vol. 66, pp. 1246–59.

Corley, CJ & Woods, AY 1991, 'Socioeconomic, sociodemographic and attitudinal correlates of the tempo of divorce', Journal of Divorce and Remarriage, vol. 16, pp. 47–68.

Craig, L 2006, 'Children and the revolution: a time-diary analysis of the impact of motherhood on daily workload', Journal of Sociology, vol. 42, pp. 125–43.

Day, LH 1964, 'Patterns of divorce in Australia and the United States', American Sociological Review, vol. 29, pp. 509–22.

De Graaf, PM & Kalmijn, M 2006, 'Change and stability in the social determinants of divorce: a comparison of marriage cohorts in the Netherlands', European Sociological Review, vol. 22, pp. 561–72.

DeMaris, A & Rao, V 1992, 'Premarital cohabitation and subsequent marital stability in the United States: a reassessment', Journal of Marriage and the Family, vol. 54, pp. 178–90.

de Vaus, DA 1997, 'Divorce', in DA de Vaus & I Wolcott (eds), Australian family profiles: social and demographic patterns, Australian Institute of Family Studies, Melbourne, pp. 25–36.

——2002, 'Marriage and mental health', Family Matters, vol. 62, pp. 26–32.

——2004, Diversity and change in Australian families: statistical profiles, Australian Institute of Family Studies, Melbourne.

de Vaus, DA, Qu, L & Weston, RE 2003, 'Premarital cohabitation and subsequent marital stability', Family Matters, vol. 65, pp. 34–39.

——2005, 'The disappearing link between premarital cohabitation and subsequent marital stability', Journal of Population Research, vol. 22, pp. 99–118.

Dush, CMK, Cohan, CL & Amato, PR 2003, 'The relationship between cohabitation and marital quality and stability: change across cohorts?', Journal of Marriage and the Family, vol. 65, pp. 539–49.

England, P & Farkas, G 1986, Households, employment, and gender, Aldine, New York.

England, P, Sayer, LC & Allison, PD 2005, 'He left, she left: gains to marriage, relative resources and divorce initiation', Population Association of America annual meeting, Philadelphia, USA.

Faulkner, RA, Davey, M & Davey, A 2005, 'Gender-related predictors of change in marital satisfaction and marital conflict', American Journal of Family Therapy, vol. 33, pp. 61–83.

Ferree, MM 1990, 'Beyond separate spheres: feminism and family research', Journal of Marriage and the Family, vol. 52, pp. 866–84.

——1991, 'The gender division of labor in two-earner households: dimensions of variability and change', Journal of Family Issues, vol. 12, pp. 158–80.

Finlay, HA 2005, To have but not to hold: a history of attitudes to marriage and divorce in Australia, The Federation Press, Sydney.

Fox, GL & Murry, VM 2000, 'Gender and families: feminist perspectives and family research', Journal of Marriage and the Family, vol. 62, pp. 1160–72.

Frisco, ML & Williams, K 2003, 'Perceived housework equity, marital happiness and divorce in dual-earner households', Journal of Family Issues, vol. 24, pp. 51–73.

Funder, K 1996, Remaking families: long-term adaptation of parents and children to divorce, Australian Institute of Family Studies, Melbourne.

Funder, K, Weston, RE & Harrison, M 1993, Settling down: pathways of parents after divorce, Australian Institute of Family Studies, Melbourne.

Gaulthier, AH, Smeeding, TM & Furstenberg, FF 2004, 'Are parents investing less time in children? Trends in selected countries', Population and Development Review, vol. 30, pp. 647–71.

Giddens, A 1992, The transformation of intimacy: sexuality, love and eroticism in modern societies, Polity Press, Cambridge.

Glezer, H 1997, 'Cohabitation and marriage relationships in the 1990s', Family Matters, vol. 47, pp. 5–9.

Gottman, JM & Notarius, CI 2002, 'Marital research in the 20th century and a research agenda for the 21st century', Family Process, vol. 41, pp. 159–97.

Greenstein, TN 1995, 'Gender ideology, marital disruption, and the employment of married women', Journal of Marriage and the Family, vol. 57, pp. 31–42.

Gross, N & Simmons, S 2002, 'Intimacy as a double-edged phenomenon? An empirical test of Giddens', Social Forces, vol. 81, pp. 531–35.

Hackstaff, K 1999, Marriage in a culture of divorce, Temple University Press, Philadelphia.

Halford, WK 2000, Australian couples in millennium three, Commonwealth of Australia, Canberra.

Halford, WK, O'Donnell, C, Lizzio, A & Wilson, KL 2006, 'Do couples at high risk of relationship problems attend premarriage education?', Journal of Family Psychology, vol. 20, pp. 160–63.

Halford, WK & Simons, M 2005, 'Couple relationship education in Australia', Family Process, vol. 44, pp. 147–59.

Hall, DR 1996, 'Marriage as a pure relationship: exploring the link between premarital cohabitation and divorce in Canada', Journal of Comparative Family Studies, vol. 27, pp. 1–12.

Hall, DR & Zhao, JZ 1995, 'Cohabitation and divorce in Canada: testing the selectivity hypothesis', Journal of Marriage and the Family, vol. 57, pp. 421–27.

Hawkins, AJ, Nock, SL, Wilson, JC, Sanchez, L & Wright, JD 2002, 'Attitudes about covenant marriage and divorce: policy implications from a three-state comparison', Family Relations, vol. 51, pp. 166–75.

Heaton, TB 1991, 'Time-related determinants of marital dissolution', Journal of Marriage and the Family, vol. 53, pp. 285–95.

Heaton, TB & Blake, AM 1999, 'Gender differences in determinants of marital disruption', Journal of Family Issues, vol. 20, pp. 25–45.

Hetherington, EM & Kelly, J 2002, For better, or for worse: divorce reconsidered, WW Norton & Company, New York.

Hewitt, B, Western, M & Baxter, J 2006, 'Who decides? The social characteristics of which spouse initiates separation', Journal of Marriage and Family, vol. 68, pp. 1165–77.

Hoem, JM 1997, 'Educational gradients in divorce risks in Sweden in recent decades', Population Studies, vol. 51, pp. 19–27.

Hoffman, SD & Duncan, GJ 1995, 'The effect of incomes, wages, and AFDC benefits on marital disruption', Journal of Human Resources, vol. 30, pp. 19–41.

Hopper, J 1993, 'The rhetoric of motives in divorce', Journal of Marriage and the Family, vol. 55, pp. 801–13.

House of Representatives Standing Committee on Family and Community Affairs (HRSCFCA) 2003, Every picture tells a story: report on the inquiry into child custody arrangements in the event of separation, HRSCFCA, Canberra.

House of Representatives Standing Committee on Legal and Constitutional Affairs (HRSCLCA) 1998, To have and to hold: strategies to strengthen marriage and relationships, HRSCLCA, Canberra.

Human Rights and Equal Opportunity Commission 2005, Striking the balance: women, men, work and family, Human Rights and Equal Opportunity Commission, Sydney.

Jalovaara, M 2003, 'The joint effects of marriage partners' socioeconomic positions on the risk of divorce', Demography, vol. 40, pp. 67–81.

Jones, FL 1994, 'Are marriages that cross ethnic boundaries more likely to end in divorce?', Journal of the Australian Population Association, vol. 11, pp. 115–32.

Kalmijn, M & Poortman, A-R 2006, 'His or her divorce? The gendered nature of divorce and its determinants', European Sociological Review, vol. 22, pp. 201–14.

Khoo, S-E & Zhao, Z 2001, 'A decomposition of immigrant divorce rates in Australia', Journal of Population Research, vol. 18, pp. 68–77.

Kiernan, KE & Cherlin, A 1999, 'Parental divorce and partnership dissolution in adulthood: evidence from a British cohort study', Population Studies, vol. 53, pp. 39–48.

Kincaid, SB & Caldwell, RA 1995, 'Marital separation: causes, coping, and consequences', Journal of Divorce and Remarriage, vol. 22, pp. 109–28.

Kitson, GC 1982, 'Attachment to the spouse in divorce: a scale and its application', Journal of Marriage and the Family, vol. 44, pp. 379–93.

Lillard, LA, Brien, MJ & Waite, LJ 1995, 'Premarital cohabitation and subsequent marital dissolution: a matter of self-selection?' Demography, vol. 32, pp. 437–57.

Lillard, LA & Waite, LJ 1995, 'Til death do us part: marital disruption and mortality', American Journal of Sociology, vol. 100, pp. 1131–56.

Lyngstad, TH 2006, 'Why do couples with highly educated parents have higher divorce rates?', European Sociological Review, vol. 22, pp. 49–60.

McDonald, P 2000, 'Gender equity, social institutions and the future of fertility', Journal of Population Research, vol. 17, pp. 1–16.

McDonald, PF & Institute of Family Studies (Australia) 1986, Settling up: property and income distribution on divorce in Australia, Prentice-Hall, Englewood Cliffs, New Jersey.

McLanahan, S & Bumpass, LL 1988, 'Intergenerational consequences of family disruption', American Journal of Sociology, vol. 94, pp. 130–52.

Markman, HJ & Halford, WK 2005, 'International perspectives on couple relationship education', Family Process, vol. 44, pp. 139–46.

Milkie, M, Bianchi, SM, Mattingly, MJ & Robinson, JP 2002, 'Gendered division of childrearing: ideals, realities, and the relationship to parental well-being', Sex Roles, vol. 47, pp. 21–38.

Moore, KA & Waite, LJ 1981, 'Marital dissolution, early motherhood and early marriage', Social Forces, vol. 60, pp. 20–40.

Morgan, PS & Rindfuss, RR 1985, 'Marital disruption: structural and temporal dimensions', American Journal of Sociology, vol. 90, pp. 1055–77.

Mueller, CW & Pope, H 1977, 'Marital instability: a study of its transmission between generations', Journal of Marriage and the Family, vol. 39, pp. 83–93.

Nock, SL, Wright, JD & Sanchez, L 1999, 'America's divorce problem', Social Science and Public Policy, vol. 36, pp. 43–52.

O'Brien, C & Rich, K 2003, Evaluation of the Men and Family Relationships Initiative: final report and supplementary report November 2002, Commonwealth Department of Family and Community Services, Canberra.

Ono, H 1998, 'Husbands' and wives' resources and marital dissolution', Journal of Marriage and the Family, vol. 60, pp. 674–89.

Oppenheimer, VK 1994, 'Women's rising employment and the future of the family in industrial societies', Population and Development Review, vol. 20, pp. 293–342.

——1997, 'Women's employment and the gain to marriage: the specialization and trading model', Annual Review of Sociology, vol. 23, pp. 431–53.

Ozdowski, SA & Hattie, J 1981, 'The impact of divorce laws on divorce rate in Australia: a time series analysis', Australian Journal of Social Issues, vol. 16, pp. 3–17.

Pettit, EJ & Bloom, BL 1984, 'Whose decision was it? The effects of initiator status on adjustment to marital disruption', Journal of Marriage and the Family, vol. 46, pp. 587–95.

Pocock, B 2005, 'Work/care regimes: institutions, culture and behaviour and the Australian case', Gender, Work and Organization, vol. 12, pp. 32–49.

Ponzetti, JJ, Zovkovic, AM, Cate, RM & Huston, TL 1992, 'Reasons for divorce: a comparison between former partners', Journal of Divorce and Remarriage, vol. 17, pp. 183–201.

Roden, M 1989, 'Covariates of divorce in Australia: an analysis using proportional hazards models', Journal of the Australian Population Association, vol. 6, pp. 145–63.

Rogers, SJ 2004, 'Dollars, dependency and divorce: four perspectives on the role of wives' income', Journal of Marriage and the Family, vol. 66, pp. 59–74.

Rogers, SJ & Amato, PR 2000, 'Have changes in gender relations affected marital quality?', Social Forces, vol. 79, pp. 731–53.

Rogers, SJ & DeBoer, DD 2001, 'Changes in wives' income: effects on marital happiness, psychological well-being, and the risk of divorce', Journal of Marriage and the Family, vol. 63, pp. 458–72.

Rosier, KB & Feld, SL 2000, 'Covenant marriage: a new alternative for traditional families', Journal of Comparative Family Studies, vol. 31, pp. 385–94.

Sanchez, L & Gager, CT 2000, 'Hard living, perceived entitlement to a great marriage and marital dissolution', Journal of Marriage and the Family, vol. 62, pp. 708–22.

Sanchez, L, Nock, SL, Wright, JD & Gager, CT 2002, 'Setting the clock forward or back? Covenant marriage and the "divorce revolution"', Journal of Family Issues, vol. 23, pp. 91–120.

Sandberg, JF & Hofferth, SL 2005, 'Changes in time with parents: a correction', Demography, vol. 42, pp. 391–95.

Sarantakos, S 1994, 'Trial cohabitation on trial', Australian Social Work, vol. 47, pp. 13–25.

Sayer, LC & Bianchi, SM 2000, 'Women's economic independence and the probability of divorce: a review and reexamination', Journal of Family Issues, vol. 21, pp. 906–43.

Sayer, LC, Binachi, S & Robinson, JP 2004, 'Are parents investing less in children? Trends in mothers' and fathers' time with children', American Journal of Sociology, vol. 110, pp. 1–43.

Schoen, R 1992, 'First unions and the stability of first marriages', Journal of Marriage and the Family, vol. 54, pp. 281–84.

Singer, JD & Willett, JB 2003, 'Applied longitudinal data analysis: modelling change and event occurrence', Oxford University Press, New York.

Smart, C 2000, 'Divorce and changing family practices in a post-traditional society: moral decline or changes to moral practices?' Family Matters, vol. 56, pp. 10–19.

Smart, C & Neale, B 1999, Family fragments? Polity Press, Cambridge.

Smyth, B 2005, 'Time to rethink time? The experiences of time with children after divorce', Family Matters, vol. 71, pp. 4–10.

Smyth, B, Caruna, C & Ferro, A 2005, 'Father–child contact after separation: profiling five different patterns of care', Family Matters, vol. 67, pp. 20–27.

Smyth, B & Weston, RE 2000, Financial living standards after divorce: a recent snapshot, Australian Institute of Family Studies, Melbourne.

South, SJ 1995, 'Do you need to shop around? Age at marriage, spousal alternatives, and marital dissolution', Journal of Family Issues, vol. 16, pp. 432–49.

——2001, 'Time-dependent effects of wives' employment on marital dissolution', American Sociological Review, vol. 66, pp. 226–45.

Stanley, SM, Amato, PR, Johnson, CA & Markman, HJ 2006, 'Premarital education, marital quality, and marital stability: findings from a large, random household survey', Journal of Family Psychology, vol. 20, pp. 117–26.

StataCorp 2003, Stata Statistical Software, Release 8.0, Stata Corporation, College Station, Texas.

Steil, JMI 1997, Marital equality: its relationship to the well-being of husbands and wives, Sage Publications, Thousand Oaks, California.

Stewart, DE & Harrison, M 1982, Divorce in Australia, Institute of Family Studies, Melbourne.

Sweeney, MM 2002, 'Remarriage and the nature of divorce: does it matter which spouse chose to leave?', Journal of Family Issues, vol. 23, pp. 410–40.

Sweeney, MM & Cancian, M 2004, 'The changing importance of white women's economic prospects for assortive mating', Journal of Marriage and the Family, vol. 66, pp. 1015–28.

Sweeney, MM & Phillips, JA 2004, 'Understanding racial differences in marital disruption: recent trends and explanations', Journal of Marriage and the Family, vol. 66, pp. 639–50.

Teachman, JD 2002a, 'Childhood living arrangements and the intergenerational transmission of divorce', Journal of Marriage and the Family, vol. 64, pp. 717–29.

——2002b, 'Stability across cohorts in divorce risk factors', Demography, vol. 39, pp. 331–51.

——2004, 'The childhood living arrangements of children and the characteristics of their marriages', Journal of Family Issues, vol. 25, pp. 86–111.

Teachman, JD & Polonko, KA 1990, 'Cohabitation and marital stability in the United States', Social Forces, vol. 69, pp. 207–20.

Teti, DM & Lamb, ME 1989, 'Socioeconomic and marital outcomes of adolescent marriage, adolescent childbirth, and their co-occurrence', Journal of Marriage and the Family, vol. 51, pp. 203–12.

Thompson, L & Walker, AJ 1989, 'Gender in families: women and men in marriage, work, and parenthood', Journal of Marriage and the Family, vol. 51, pp. 845–71.

——1995, 'The place of feminism in family studies', Journal of Marriage and the Family, vol. 57, pp. 847–65.

Thornton, A 1977, 'Children and marital stability', Journal of Marriage and the Family, vol. 39, pp. 531–40.

Thornton, A & Rodgers, WL 1987, 'The influence of individual and historical time on marital dissolution', Demography, vol. 24, pp. 1–22.

Tzeng, JM & Mare, RD 1995, 'Labor market and socioeconomic effects on marital stability', Social Science Research, vol. 24, pp. 329–51.

Vannoy, D 2000a, 'Roles in the divorce process and identity strength', Journal of Divorce and Remarriage, vol. 32, pp. 101–18.

——2000b, 'Roles in the divorce process: marital quality and relationship intimacy: a research note', Journal of Divorce and Remarriage, vol. 32, pp. 119–24.

Waite, LJ & Lehrer, EL 2003, 'The benefits from marriage and religion in the United States: a comparative analysis', Population and Development Review, vol. 29, pp. 255–75.

Waite, LJ & Lillard, LA 1991, 'Children and marital disruption', American Journal of Sociology, vol. 96, pp. 930–53.

Walzer, S & Oles, TP 2003, 'Accounting for divorce: gender and uncoupling narratives', Qualitative Sociology, vol. 26, pp. 331–49.

Wang, H & Amato, PR 2000, 'Predictors of divorce adjustment: stressors, resources, and definitions', Journal of Marriage and the Family, vol. 62, pp. 655–68.

Ware, H 1975, 'Immigrant fertility: behaviour and attitudes', International Migration Review, vol. 9, pp. 361–78.

Watson, N & Wooden, M 2002a, 'Assessing the quality of the HILDA Survey Wave 1 data', Australian Government Department of Family and Community Services & The University of Melbourne, Melbourne.

——2002b, 'The Household, Income and Labour Dynamics in Australia (HILDA) survey: Wave 1 survey methodology', The University of Melbourne.

Western, M, Baxter, J & Chesters, J 2007, 'Australian families: just managing', in D Denemark, G Meagher, S Wilson, T Phillips & M Western (eds), The Australian survey of social attitudes: the second report, The University of New South Wales Press, Sydney, pp. 241–61.

White, LK 1990, 'Determinants of divorce: a review of research in the eighties', Journal of Marriage and the Family, vol. 52, pp. 904–12.

Wolcott, I & Hughes, J 1999, Towards understanding the reasons for divorce, Australian Institute of Family Studies, Melbourne.

Wolfinger, NH 1999, 'Trends in the intergenerational transmission of divorce', Demography, vol. 36, pp. 415–20.

——2000, 'Beyond the intergenerational transmission of divorce: do people replicate the patterns of marital instability they grew up with?' Journal of Family Issues, vol. 21, pp. 1061–86.

Yodanis, C 2005, 'Divorce culture and marital gender equality: a cross-national study', Gender and Society, vol. 19, pp. 644–59.

[ Return to Top   Return to Section ]

List of tables and figures

Table 1: Distribution of separated and divorced in analytic sample

Table 2: Description of covariates, by sex

Table 3: Discrete time event history model predicting the risk of marriage breakdown as a function of various social characteristics(a) for men and women

Table 4: Initiator status of separation for separated and divorced men and women

Table 5: Discrete time event history models with competing risks of women's reports of who initiated separation relative to still married by social characteristics

Table 6: Discrete time event history models with competing risks of men's reports of who initiated separation relative to still married by social characteristics

Table 7: Competing risks model, contrasting wife-initiated separation by selected social characteristics for women and men

Appendix tables

Table A1: Likelihood ratio tests comparing model in Table 3 with models excluding each of the covariates

Table B1: Gender interactions for discrete time event history model presented in Table 3

Table C1: Social characteristics interacted with gender by reports of who initiated separation

List of figures

Figure 1: Survival estimates of marriage breakdown for analytic sample, by sex

Figure 2: Hazard rate of marriage breakdown by duration of marriage in analytic sample, by sex

Figure 3: Hazard rates of marriage breakdown and men's and women's reports of who initiated separation

[ Return to Top   Return to Section ]

Last updated: