Note: Where available, the PDF/Word icon below is provided to view the complete and fully formatted document
Pathways from cohabitation. Paper presented at the HILDA Survey Research Conference, University of Melbourne, September 29-30, 2005



Download PDFDownload PDF

Pathways From Cohabitation

Ruth Weston

Australian Institute of Family Studies

Lixia Qu

Australian Institute of Family Studies

David de Vaus

School of Social Sciences, La Trobe University

Paper presented at the HILDA Survey Research Conference, University of

Melbourne, September 29-30, 2005

1

Pathways from cohabitation

Weston, R., Qu, L., and de Vaus, D.

ABSTRACT

This study seeks to identify the factors that discriminate between

cohabiting couples who choose to marry, those who choose to

separate, and those who continue to cohabit. Three waves of

HILDA data are used to explore the characteristics of those who

follow these different pathways. These data may help explain why

many cohabitors marry in an era when cohabitation is common and

on the rise, and when children are increasingly born outside

marriages. There is also a voluminous literature on divorce, but a

patchy and scant literature on cohabiting relationships that break

down.

History is punctuated by periods of rapid change and unpredictability brought about

through revolutions, wars, economic downturns and other social upheavals, and

interspersed between periods of relative stability. Today’s world is a case in point:

rapid change continues along many interconnected fronts, including the labour market

and economy, technology, personal relationships, and social attitudes. Many

permanent jobs are being replaced with fixed-term contracts or “exported” to

developing countries, and in many professions, fast and accelerating technological

advancements are making redundant years of training and experience. Personal

relationships have become increasingly unstable, and some of the social constraints

which shaped the way in which people live their lives have weakened, while others

have disappeared.

In an uncertain world, where jobs, relationships and financial resources may seem

tenuous, people may respond by treading cautiously before making commitments that

involve heavy and long-term financial obligations, such as marrying and having

children. One strategy is to live with a partner first (here called “cohabitation”) to test

each other’s compatibility before making any decision about marriage and starting a

family. It may also be the case that other preconditions have to be met before people

will marry. For instance, couples may reduce the risk of marriage breakdown by

ensuring not just that they continue to enjoy their relationship and that they agree on

important issues such as having children, but also that they will be financially secure.

There are, of course, several other reasons for cohabiting. Some couples who are

already committed to marrying may live together first to save money, while others

may view marriage as “just a piece of paper”. In addition, some couples may move in

together for the adventure, taking each day as it comes, and some may consider the

lifestyle as one of “no strings attached”. Of course, the meaning of cohabitation can

change, and each partner may interpret the relationship differently.

Whatever its meaning, cohabitation has become an increasingly prevalent part of the

life course and the course it takes now varies. These trends are outlined below and

are followed by a discussion of previous research into the correlates of the different

trajectories.

2

Changing patterns of partnership formation

While the vast majority of couples who live together are married to each other, the

proportion who are cohabiting increased from less than one per cent in 1971 and

nearly 5 per cent in 1986 to 12 per cent in 2001 (Carmichael 1995; Santow and

Bracher 1994, ABS 2001). Furthermore, cohabitation is the most common living-

together arrangement of those under the age of 25 (applying to 82 per cent of

partnered individuals under 20 years old, and 61 per cent of partnered individuals

aged 20-24 years) (de Vaus 2004).

The greater tendency for younger than older couples to cohabit reflects the recency of

this living arrangement and the fact that it has become the normative pathway to

marriage (applying to around 75 per cent of all couples who married in 2003,

compared with only 16 per cent in 1975 (ABS 2004a)).

Furthermore, the premarital cohabitation period has itself lengthened, thereby

contributing to the prevalence of cohabitation amongst young adults. Around 40 per

cent of couples who experienced premarital cohabitation in the 1960s or 1970s

married within six months, compared with only 12 per cent who experienced

premarital cohabitation in the late 1990s. Conversely, the proportions who lived

together for at least three years before marrying increased from 5 per cent in the

1960s or 1970s to 30 per cent in the late 1990s (Weston, de Vaus and Qu 2003).

The instability of cohabitation

Despite the increase in cohabitation and the lengthening of its period for those who

marry, in the vast majority of cases, cohabitation is relatively short-lived. It appears

that only nine per cent of those whose cohabitation commenced in the early 1990s

were still cohabiting with the same partner in 2001 (7‐11 years later). Although

cohabitation is now the normative pathway to marriage, the chance of it ending in

separation rather than marriage has increased. Of those whose first unions

commenced in the early 1970s, 63 per cent had married within five years and 25 per

cent had separated. On the other hand, of those who commenced cohabiting in the

early 1990s, the chance of separation or marriage occurring within five years was

similar (43 per cent and 38 per cent respectively) (Weston et al. 2003, de Vaus 2004).

What factors are linked with these different pathways from cohabitation?

Correlates of cohabitation pathways

While a great deal of research attention has been directed to the key correlates of

marriage breakdown and of premarital cohabitation (as opposed to direct marriage)

(see Wolcott and Hughes 1999; de Vaus, Weston and Qu, 2003a), as cohabitation has

increased in prevalence, there has been growing research interest in identifying the

factors that influence, or are at least associated with, cohabitation transitions. The

following factors are relevant to the present paper: the couple’s financial

circumstances and related socio-demographic characteristics, the quality of their

relationship, the duration of their cohabitation, the partners’ ages, experience of

previous relationships, and family type. Findings on these issues are outlined below.

3

Financial circumstances. Consistent with the above argument that people are now

very cautious about committing to marriage, overseas research (most commonly but

not exclusively in the US) suggests that recent cohorts of cohabiting couples tend to

defer marriage until they believe that they are in a financially secure position (Brown

2000; Cherlin 2004; Duvander 2000; Ermisch and Francesconi 2000; Smock and

Manning 1995, 1997; Smock, Manning and Porter 2005). Cherlin linked this trend

with changes in the meaning of marriage. In his view, now that marriage is no longer

the only socially sanctioned arrangement for sexual relationships and having children,

it has become a symbol of having achieved a relationship that is both emotionally

rewarding and financially secure. According to Cherlin, many couples with limited

financial prospects postpone marriage indefinitely as they strive unsuccessfully to

attain such security. In short, he argued that “People marry now less for the social

benefits that marriage provides than for the personal achievements it represents”

(p.857).

While overseas research suggests that financial wellbeing increases the probability of

marriage, the impact of financial difficulties on cohabitating relationships remains

unclear (see Smock et al. 2005) ‐ even though financial pressures appear to increase

the risks of marriage breakdown (see Wolcott and Hughes 1999).

Overseas research also suggests that the transition from cohabitation to marriage is

contingent on the male partner’s “financial characteristics”, as measured by education,

occupation or earnings (for a review, see Smock et al. 2005). These findings are

consistent with Becker’s (1981) argument that marriage is attractive to partners when

each can offer a complementary rather than competitive role. The traditional

complementary role is that of a male partner who is a “good provider” and a female

partner who is a “good homemaker”. Despite the increase in dual earner families

since Becker published these arguments

1

, a continuing emphasis on the male partner’s

financial characteristics is understandable given that the strength of women’s (but not

men’s) attachment to the labour force often depends on the age of their children

should they become parents.

But does the impact of the male partner’s financial characteristics vary according to

the female partner’s financial characteristics? Oppenheimer (1988) argued that, as

women have increasingly been expected to remain attached to the labour force and as

men’s ability to support a family alone has diminished, women’s earnings have

become increasingly important in decisions about marriage, while men’s earnings

have diminished in importance. Becker, on the other hand, maintained that the

attractiveness of marriage to women is reduced by their personal economic

independence, while Smock and Manning (1997) proposed that a woman’s strong

economic position undermines her need for marriage while also making her an

attractive marriage partner.

So far, studies that have examined the impact of women’s educational level or

earnings have yielded mixed results (see Smock et al. 2005). However, it is important

to note that most of the research in this area has focused independently on the

economic-related characteristics of men and/or women rather than on the relative

1

For recent Australian trends, see Gray, Qu. Renda and de Vaus (2003). See also Baxter (2005) for an

analysis of the work transition patterns of mothers linked with childbirth and factors linked with these

patterns.

4

characteristics of each partner. A key problem is that many studies only have

information on one of the partners ‐ an impediment that can lead to misleading

conclusions (see Smock and Manning 1997; see Smock et al. 2005).

Perceived relationship quality. While economic circumstances appear to be

important factors influencing union transitions, the quality of the relationship should

be crucial. As noted above, many couples may see cohabitation as a “trial marriage” ‐

an opportunity to gain unique insight into their compatibility, thereby helping them to

make informed decisions about whether or not to marry. To the extent that

cohabitation is a “trial marriage”, couples who are very happy with the quality of their

relationship might be expected to marry.

Brown’s (2000) US study, which was based on couple data, suggested that the impact

of cohabiting couples’ evaluations of their relationship on union transitions is not

straightforward. Firstly, relationship assessments appeared to have a greater bearing

on separation than marriage: the odds of separation increased with negative

assessments and decreased with positive assessments, while positive assessments did

not appear to influence the odds of marriage. Secondly, negative assessments

provided by the female partner increased the odds of separation, while such an

assessment by the male partner decreased the odds of marriage. Thus, continuation of

the relationship seemed more contingent on the female than male being happy with

the relationship. The latter finding is consistent with research in Australia on marriage

breakdown: divorced men and women are inclined to report that it was the former

wife who initiated the separation decision (e.g. Harrison 1986; Wolcott and Hughes

2000; de Vaus 2004).

Marriage expectations. Some partners who are highly satisfied with their relationship

may prefer not to marry, and others who want marriage may not expect it. For

instance, they might know that their partner is against marriage, or they or their

partner may have established preconditions for marriage that seem impossible to

achieve, as outlined by Cherlin (2004). Nevertheless, it is not surprising that previous

research suggests that cohabitors who expect to marry are much more likely to do so

than others, while couples in which neither partner plans to marry have an increased

probability of separating rather than continuing to cohabit (see Brown 2000).

Age. Age is also likely to affect cohabitation pathways. In Canada, Wu and

Balakrishnan (1995) found that age at the start of the union was negatively related to

both separation and marriage, while in the US, Brown (2000) found this relationship

for marriage but not separation. However, the timing of marriage is likely to be

affected by the partners’ current age, for marriage tends to peak at age 25‐29 for

women and at age 25‐34 for men. It therefore seems likely that the age of cohabiting

partners will have a non-linear effect on transitions from cohabitation to marriage.

The age gap between partners may also influence pathways, but the nature of any

such change is unclear. Shehan, Berardo,Vera and Carley (1991) pointed out that

partners of similar ages tend to share similar experiences, attitudes and behavioural

patterns. A large age gap may reflect lower compatibility in these areas, thereby

5

increasing the chance of conflict and separation. However, Wu and Balakrishnan

found that couples were less likely to separate and more likely to marry when the

male partner was at least 10 years older than the female partner. On the other hand,

Tzeng (1992) found that, among married couples, the chance of separation increased

when the husband was more than three years older than his wife. In both these

studies, no significant effects emerged where the female partner was older than the

male partner. But Lyngstad (2004) found that age heterogamy between married

spouses in Norway increased the probability of divorce, especially if the female

partner was older than the male partner.

Relationship duration. The decision to separate or marry is likely to occur in the first

few years of the union, although as noted earlier, the period of pre-marital

cohabitation has lengthened. Relationship difficulties are likely to loom large soon

after the couple moves in together, and many couples may decide to “cut their losses”

and separate before investing much time in the relationship. Furthermore, the longer

that the union has already existed, the more likely it is being seen by the couple as an

alternative to marriage, rather than as a trial marriage. Consistent with these

arguments, Brown’s (2000) analysis of data from a US follow-up study indicated that

the longer the couples had been cohabiting by wave 1, the lower were their chances of

marrying or separating. Nevertheless, it seems reasonable to suggest that the effects of

union duration on transition to marriage or separation would diminish progressively

with increasing length of the relationship.

Family type and desire for children. With so little understanding of the meaning of

cohabitation to couples, it is hard to tell what dynamics may be at play in individual

scenarios. For instance, those who view marriage as the only appropriate context for

having children would be likely to marry if they are childless and want children, or if

they had recently had their first child or achieved conception. Those who have had

children for some time while cohabiting may well see cohabitation as an alternative to

marriage.

Regardless of whether they decide to marry, there is evidence that children born of the

union lower the risks of separation (White 1990; Wu 1995). On the other hand, step

relationships appear to add pressure on unions. While remarriages are more likely

than first marriages to end in divorce (Coleman, Ganong and Fine 2000; de Vaus

1997), US research suggests that this is particularly the case for remarriages with step

children (Booth and Edwards 1992).

Experience of previous relationships. Several reasons have been proposed for the

higher instability of remarriages compared with first marriages (see Coleman et al).

For example, compared with marriages that are the first for both spouses, the

population of remarriages would contain a higher proportion of spouses with

characteristics that increase the risk of marriage breakdown, and which contributed to

the instability of their first marriage (e.g., they may exhibit behaviour, attitudes or

interests that would tend to threaten their marriage, or they may be particularly

intolerant of idiosyncrasies in others). Remarriages may also include a relatively high

proportion of spouses who see separation as a solution to problems in the relationship,

6

or the experience of a previous divorce may lead some people to be more ready to cut

their losses when problems arise. Much the same reasons would suggest that

“previous cohabitations” (i.e. where one or both partners has experienced separation

rather than marriage from a previous cohabiting relationship) would be more likely

than “first cohabitations” to end in separation.

Nevertheless, in the studies by Brown (2000), Smock and Manning (1997) and Wu

and Balakrishnan (1995), the probability of separation was not significantly affected

by a previous marriage experienced by either the male or female partner (taken

separately). Similarly, Brown found no significant change in the probability of

separation where a previous cohabitation was experienced by either partner.

Screening in of unsuitable matches? So far, the discussion has implied that

cohabitation is often treated as a “trial marriage” and thus as a means of screening out

unsuitable matches. However, research results have not always been consistent with

this hypothesis. For example, Brown (2000) failed to observe an increase in the

probability of marriage when both partners were happy with the relationship. It is

possible that “screening in” of unsuitable matches sometimes occur. For instance,

some couples may embark on marriage as a means of firming up a relationship that

may be wavering or lacking in certain desired qualities. It would also make sense that

the more time that someone has spent in the relationship, the greater the urge to

maintain it. Furthermore, older cohabitors who want marriage may feel that they have

little prospect of finding a more suitable partner than their current partner. The same

may apply for those who feel they lack the qualities that make for an attractive

marriage partner.

The present study traces the relationship pathways of Australian couples who were

cohabiting when first surveyed. It examines the extent to which factors outlined

above, such as each partner’s satisfaction with the relationship, predict the pathways

taken. As an adjunct, it also examines circumstances that might suggest a “screening

into” marriage of apparently unsuitable matches.

Sample and design of analysis

This study is based on the first three (annual) waves of data collected in the

Household, Income and Labour Dynamics in Australia (HILDA) Survey.

2

A nation-

wide sample of private households was selected in the first wave (which took place in

2001), and interviews and self-completed questionnaires were sought from household

members aged 15 years or more. In total, information was collected from nearly

14,000 people in close to 7,700 households in wave 1. All members of these

households are treated as “continuing household members”. Subsequent waves

involve attempts to trace these individuals and interview all those aged 15 years or

2

This survey is funded by the Australian Government through the Department of Family and

Community Services. It is managed by a consortium led by the Melbourne Institute of Applied

Economic and Social Research. The other partners are the Australian Institute of Family Studies and

the Australian Council for Educational Research. Details of the survey design are provided by Watson

and Wooden (2002). The questionnaires used in each wave and further details of the survey can be

found at the web address, http:// www.melbourneinstitute.com/hilda/

7

more, along with any “new” people of this age in their household. A self-completed

questionnaire is also administered to each continuing and new sample member.

3

The analysis outlined below focuses on the pathways that were followed over a two-

year period (marriage, separation, or no change) by heterosexual couples who were

cohabiting in wave 1, and the factors linked with each pathway. Couple-level

measures were used instead of individual measures. As Brown (2000) pointed out,

pathways are the result of a joint decision and the use of individual measures

regarding relationship assessments and expectations can lead to biased results.

Furthermore, couple-level measures make it possible to examine the role of gender in

union transitions.

In total, there were 714 cohabiting heterosexual couples in wave1.

4

At that time,

interviews with each partner were conducted for 593 of these couples and with only

one partner for 122 couples. Nevertheless, socio-demographic characteristics such as

age, employment status, and the nature of their relationship to all other household

members, were ascertained for all members of the household, regardless of whether

they were interviewed.

The first set of analysis, which focuses on the pathways taken by cohabiting couples,

is based on all 714 couples. As shown later, not all couples were interviewed in all

three waves. Attention is then directed towards relationships between selected socio-

demographic and subjective factors and the pathways followed by couples. This

analysis is based on couples in which each partner was aged between 20 and 55 years

in wave 1. This age range was employed to allow for a reasonable chance of marriage

taking place.

5

Unless otherwise specified, the data used in this analysis are those that were derived

in the wave preceding any marriage or separation. For example, for couples who had

married or separated by wave 2, the information they provided in wave 1 (e.g.

regarding relationship satisfaction) was used, while for those who married or

separated between waves 2 and 3, the information they provided in wave 2 was used.

Data derived in wave 2 were used for couples who were still cohabiting (involving the

same partners) in wave 3.

The analysis of the relationship between variables and cohabitation pathways

involved two stages. Bivariate analysis was first conducted and the results helped

guide the selection of measures used in the second stage, which involved application

of multinomial logistic regression.

3

New partners become continuing sample members if they have a child with a continuing sample

member; the children born of this relationship themselves become continuing sample members.

4

There were 22 same-sex couples. This group was excluded on the grounds that marriage in Australia

is not an option for these couples.

5

In 2001, the marriage rates were 0.9 males and 4.2 females per 1000 unmarried male and female

teenagers respectively in 2001. The rates for those in their early twenties were 23.1 and 42.6 per 1000

unmarried men and women respectively aged 15 years and over, while those for men and women in

their late forties were 28.1 and 22.4 respectively. However, the marriage rate for women aged 50 and

over was considerably lower than the latter rate (4.6 per 1000), that for males was 14.4 per 1000 (ABS

2002). The rate for those in their early fifties would, of course, be higher than those provided by the

ABS which cover all individuals aged 50 years and over.

8

Measures and hypotheses

The predictors of union transitions examined in this study comprised economic

circumstances, relationship satisfaction, desire for children, family type, history of

relationships, the length of current relationship, and the female partner’s age. While

the precise measures are listed in Table 1, most require further explanation.

Economic circumstances.

As noted above, several studies suggest that a couple’s economic circumstances are

important in predicting outcomes of cohabitation, although the impact of relative

differences between the male and female partner’s economic characteristics remains

unclear. There is considerable evidence that overall financial wellbeing and men’s

financial resources or earning potential increase the odds of marriage, but the impact

of women’s economic characteristics remains uncertain.

In the present study, proxies for the couple’s overall financial wellbeing and earning

capacity of one partner relative to the other partner were used. These covered

educational attainment, engagement in study, employment status, income and self-

assessed financial circumstances.

The educational attainment measure was used as a proxy for earning capacity and

focused on whether both partners, the male only, the female only, or neither partner

held a degree. Overall financial wellbeing was assessed by the combined annual

incomes of each partner (and reduced to three categories based on the overall

distribution derived for this measure), while perceived financial wellbeing was based

on each partner’s answer to the question: “Given your current needs and financial

responsibilities, would you say that you and your family were prosperous, very

comfortable, reasonably comfortable, just getting along, poor, or very poor?” Given

the distribution of responses, four categories were derived. These indicated whether

both partners, the male only, the female only, or neither partner reported that they

were reasonably comfortable, very comfortable or prosperous (here called

“comfortable”).

6

This subjective measure was included in the self-completion

questionnaire, and in 12 per cent of cases one or both partners either skipped this

question or did not complete the entire questionnaire. To prevent loss of all other data

for couples who did not complete this question, a “not stated” category was created

for the subjective measure of financial wellbeing. The employment status measure

focused on which partner if any was working full-time: both, male partner only,

female partner only, and neither.

On the basis of previous research and the argument that marriage is more likely when

the risks of marriage breakdown are regarded as having been minimised, it was

predicted that:

H

1

: couples whose combined incomes were relatively high and those in which

both partners agreed that they were in a comfortable or better financial

position would be most likely to marry, while separation would be most

6

Nearly 80 per cent of male partners reported that they were “reasonably comfortable” or “just getting

along”, while less than 2 per cent indicated that they were “poor” or “very poor”.

9

common amongst couples with low combined incomes and amongst

couples in which both partners agreed that they were not in a financially

comfortable position.

Furthermore, it was predicted that:

H

2

:couples in which the male partner held a degree would be more likely than

other couples to marry, while those in which neither partner held a degree

would be the most likely of all couples to separate.

In addition, given the emphasis couples appear to place on "saving up" to marry (see

Cherlin 2004), and the likelihood that part-time work for a male partner is not chosen

(when the effects of his possibly being a student are controlled), it was predicted that:

H

3

: marriage would be particularly likely if both partners were working full-

time, while separation would be particularly likely if the male partner was

not working full-time.

Finally, given that people who are engaged in study tend to be relatively young, their

study commitments would typically limit their work hours and therefore income, it

was predicted that:

H

4

: partners would be unlikely to marry if one or both of them were engaged

in study.

Relationship satisfaction

Brown’s (2000) research in the US suggested that, while negative evaluations of the

relationship increased the chances of separation, shared positive evaluations of the

relationship impeded separation but did not increase the odds of marriage. However,

Brown’s analysis was based on a follow-up study in the US that was conducted some

time ago, and involved around five years between waves (1987-88 and 1992-94).

In part contradiction of Brown’s findings, it predicted that:

H

5

: couples in which both partners indicated high satisfaction with their

relationship would be more likely than other couples to marry, and couples

in which neither partner indicated high relationship satisfaction would be

more likely than others to separate.

In HILDA, each partner was asked to rate their level of satisfaction on an 11-point

scale ranging from (0) “completely dissatisfied” to (10) “completely satisfied”. More

than half the male and female partners provided ratings of 9 or 10, here classified as

“high satisfaction”. Four categories of relative satisfaction were derived, indicating

which partner, if any, was highly satisfied: both, neither, male partner only, or female

partner only. This question was included in the self-completion questionnaire, and

data are missing for 12 per cent of couples. A “not stated” category was created in

order to maintain the sample size.

Screening in of unsuitable matches?

It was argued above that some couples may decide to marry when not particularly

happy with the relationship (e.g. if they have reached an age when the pool of suitable

partners is small and if the couple had already invested considerable time in the

10

relationship). Or they may attempt to “cement” a relationship that may not be

particularly happy by getting married (a “screening in” process). Again, this may

apply at older ages when opportunities of finding a more suitable partner for marriage

seem slight and when the couple had already spent a considerable time together.

It was therefore predicted that:

H

6

: amongst those who married despite one or both partners having lacked

high relationship satisfaction, at least one partner would tend to be older

and the couple would have lived together for a longer period of time,

compared with others who married.

Nevertheless, it was also predicted that:

H

7

: this “relationship rescuing” hypothesis would only weaken but not prevent

the occurrence of a strong positive relationship between satisfaction and

marriage. That is, it was predicted that relatively high relationship

satisfaction shared by each partner would increase the probability of

marriage, while lack of high satisfaction on the part of each partner would

increase the probability separation.

Marriage expectations

For many couples, a shared expectation of marriage may involve mutual active

planning for marriage. On the other hand, couples may share the expectation that no

marriage will take place if at least one partner indicates uncertainty or disillusionment

about the relationship or rejection of marriage on ideological grounds.

It was therefore predicted that:

H

8

: couples in which partners shared the expectation that they would marry

would be more likely to marry while those who shared the expectation that

they would not marry would be more likely than others to separate.

Respondents indicated whether they considered that marriage was “very likely”,

“likely”, “not sure”, “unlikely” or “very unlikely”. Three categories of conditions

were derived from these responses: both partners expected to marry (that is, both

considered marriage to be likely or very likely”; only one partner expected to marry;

and neither partner expected to marry.

Desire for a child or more children

As noted above, links between fertility aspirations and marriage or continuing

cohabitation are likely to vary for those who believe that children should be raised in

marriage and for those who see cohabitation as an alternative to marriage and thus as

a suitable for context for childbearing. Such information about cohabitation and

marriage has not been derived in HILDA. Nevertheless, given that most children are

born within marriage (68 per cent in 2003) (ABS 2004b), it seems likely that most

individuals who have a strong desire for children would like also to marry.

Furthermore, if one partner very much wants children and the other does not, then

separation may occur.

During the personal interview, partnered respondents were asked to rate their feelings

about having a child or more children on an 11-point scale ranging from 0 “Would

11

definitely not like to have a child/more children” to 10 “Would very much like to

have a child/more children”. The most common responses were either 0 or 10

(applying to one third and one quarter of partnered respondents respectively).

Responses of 9 or 10 are here considered to represent a strong desire for a child.

It was predicted that

H

9

: a shared strong desire to have children would increase the probability of

marriage, while separation would be particularly likely where one partner

wanted children and the other did not.

Family type

As already noted, cohabitation can have several meanings. Some may use it as a trial

marriage, some may be saving up to marry, and some may decide to marry if and

when they want children. Others may view cohabitation as an alternative to marriage

or as a convenient arrangement involving little if any commitment. Furthermore, each

partner may hold different interpretations and their views may change as the

arrangement unfolds. To some extent these interpretations may be reflected in family

type.

A somewhat complex measure of family type was created in order to take into

account the possibility that a first pregnancy or birth may encourage marriage. The

measure combined family type at wave 1, with information about whether a

pregnancy or birth between the same waves that any transition took place (or between

waves 2 and 3 if no transition occurred).

Five categories were derived, two of which entailed an absence of children in the

wave prior to any transition (or in wave 2 if no transition took place). Given that

pregnancy or a recent first birth may spur marriage, one of these categories

encompassed cases where a conception or birth had occurred by the next wave (when

the relationship transition became apparent, or by wave 3 if no relationship transition

occurred. The remaining three categories covered intact families where no conception

or birth between waves occurred; step or blended families where no conception of

birth between waves occurred; and “other families”.

It was predicted that

H

10

: the probability of marriage would be greatest amongst couples with no

children in the household, and amongst such “childless” couples who

experienced a recent pregnancy or first birth.

It was also predicted that

H

11

: the chance of marriage would be lowered if the couple already had

children together (apart from a recent first birth), for the presence of such

children may signify the view by one or both partners that cohabitation is

no different from marriage.

Finally, given the additional pressures that families with stepchildren and the

increased risk of marriage breakdown experienced by stepfamilies (see Coleman et al.

2000), it was predicted that:

12

H

12

: the probability of separation would be increased if there were

stepchildren in the household.

Female partner’s age

Given that marriage rates vary across age groups, peaking at 25-29 years for women

and 25‐34 years for men then declining progressively, it was predicted that:

H

13

: a similar non-linear effect of age on the transition from cohabitation to

marriage would be apparent; older cohabitators would be more likely to

marry than younger cohabitators, but the effect would diminish with

increasing age.

Because the male partner’s age was highly correlated with that of the female partner (r

=0.8), only the female partner’s age is used in the model. In the multinomial logistic

regression, this non-linear effect was tested by including as a predictor the square of

the female partner’s age.

Relative difference in ages

Despite the inconsistent results emerging from research, arguments in the literature

suggesting that a large age gap between partners may create tension lead to the

following prediction:

H

14

: the chance of separation would increase and the chance of marriage

would decrease where the male was at least 10 years older or at least 5

years younger than the female partner.

These age differences were chosen because they are non-normative but nonetheless

included at least 30 couples in each group.

Duration of cohabitation by wave 1

On the whole, it was expected that partners would typically become aware of any

signs of key areas of incompatibility fairly early in their union and would would “cut

their losses” and separate before investing much more time in the relationship.

In the present analysis, the relationship duration refers to how long the couple had

been cohabiting when interviewed in wave 1. It was predicted that:

H

15

: the chances of separation would be higher amongst those who had already

cohabited for relatively short periods of time but that this effect would be

reduced as length of cohabitation increased.

That is, a curvilinear relationship between duration of relationship and the probability

of separation was expected. The existence of a non-linear effect was tested by using

the square of relationship duration in the multivariate model.

History of relationships

It was expected that the experience of a previous marriage or cohabitation would

increase the likelihood of instability of further unions for a number of reasons. For

example: the person may possess some unattractive qualities that threaten

13

relationships; he or she may be highly sensitive to, or intolerant of, problems in

relationships; and the experience of previous separation may spur thoughts of ending

a current relationship soon after any problems in it become apparent.

Two measures of previous relationships were used: whether one or both partners had

ever been married, and whether one or both partners had experienced a previous

cohabitation that did not lead to marriage.

It was predicted that:

H

16

: the chances of separation would increase if at least one partner in the

couple had previously cohabited or married.

Results

Figure 1 shows that, by Wave 3, 23 per cent of the sample (n = 164) were no longer

interviewed. Of these, 60 per cent (n = 98 ) refused to participate and 17 per cent (n

=27) could not be traced. The other couples who were not interviewed by wave 3

were either away during the interview period, or were unable to participate for various

reasons such as poor health.

Most commonly, the couples in wave 1 were still cohabiting by 2003 (46 per cent of

all couples including those not interviewed in 2003), while a slightly higher

proportion had separated (17 per cent) rather than married (14 per cent). Excluding

those for whom no information was available by wave 3, 60 per cent were still

cohabiting in wave 3, 22 per cent had separated, and 18 per cent had married. (Of the

9 per cent of unions that converted to marriage by wave 2, only one couple had

separated by wave 3.) For nearly one-third of couples who had separated by wave 2,

at least one partner had re-partnered by wave 3.

7

Figure 2 shows that, for those who continued in the study until wave 3, any transition

was more likely to have occurred before rather than after wave 2. This trend was

slightly more apparent for separation than marriage (separation: 15 per cent versus 7

per cent; marriage: 11 per cent versus 7 per cent).

On average, the sample studied had been cohabiting for 5.3 years when the HILDA

survey was initiated. Those who were still cohabiting by wave 3 had been living

together for an average of 6.2 years when first interviewed, while those who

subsequently married or separated had been cohabiting for 3.8 and 4.1 years

respectively. What factors appear to be linked with these different pathways? The

remaining analysis focuses on 398 couples, based on information provided by 473

women and 431 men.

Bivariate analysis

This section examines the bivariate relationships between the above-noted couple-

level characteristics and cohabitation pathways (Table 1). The analysis focuses on

7

Using the British Household Panel Survey data, Ermisch (2002) found that, of respondents who had

separated from a cohabiting relationship, one-third had re-partnered within in a year of the separation

and half had separated within two years.

14

398 couples in which both partners were aged 20 to 55 years (based on information

provided by 473 women and 431 men).

It seems worth reiterating that, given that circumstances may have changed just prior

to any transition in relationship status, the measures used almost exclusively represent

those derived immediately prior to any transition (and in wave 2 if no transition had

taken place at all). The exception concerned the family type measure which took into

account circumstances in wave 1 and any conception of birth or a child that occurred

between the waves during which marriage or separation also occurred (or between

waves 2 and 3 if the couple continued to cohabit to wave 3).

Contrary to predictions, cohabitation pathways did not vary significantly according to

whether one or both partners was a student (H

4

), had ever been married or had ever

cohabitated previously (where no marriage took place) (H

16

), or according to the size

of the age gap between the partners (H

14

). It was expected that couples in which at

least one partner had been in a previous relationship (marriage or cohabitation) and

those with a marked age gap between partners would be less likely to marry and more

likely to separate than other couples. It was also expected that students would be less

likely than others to marry.

Most of the other predictions were supported in the bivariate analysis. Overall, 20 per

cent of the wave 1 cohabiting couples who remained in the panel had married by

wave 3. Table 1 shows that couples who were most likely to marry had the following

characteristics: the male partner alone held a degree (47 per cent), the couple had a

high combined income (27 per cent), both partners worked full-time (35 per cent),

both believed that they were in a financially comfortable (or better) position (30 per

cent), both were very satisfied with their relationship (36 per cent), both expected to

marry (35 per cent), both wanted a child or more children (28 per cent), and the

couple had a recent pregnancy or birth and no other children in the household (45 per

cent).

Conversely, marriage was relatively unlikely among couples in which the female

partner was not employed full-time (10‐11 per cent), the female partner did not

express high satisfaction with the relationship (including where neither partner was

satisfied) (7 to 14 per cent), only one or neither partner expected to marry (4‐5 per

cent), and the couple already had children of their union or a step or blended family in

wave 1 (9‐10 per cent). It is interesting to note that the couple seemed more likely to

marry if the female partner alone, rather than the male partner alone, provided a very

favourable view of their financial circumstances (21 per cent versus 14 per cent) or

relationship (27 per cent versus 14 per cent). This may suggest that marriage was

more contingent on the female than male partner’s evaluation of the relationship,

including its financial aspects.

In total, 19 per cent of couples had separated by wave 3. Those most likely to do so

were: couples in which there was not agreement about wanting a child (27 and 35 per

cent), and couples who had a step or blended family in wave 1 (31 per cent). Those

least likely to separate were couples in which the male partner alone held a degree (13

per cent), both partners considered that they were in a comfortable or better financial

position (13 per cent), both were very satisfied with the relationship (11 per cent),

both expected to marry (14 per cent) and both wanted children (13 per cent). In other

15

words, those who were least likely to separate were those with the same

characteristics as those who were likely to marry.

Overall, 61 per cent of the couples were still cohabiting by wave 3. The couples who

were most likely to continue to cohabit were those in which: the female partner was

not working full time (63 and 72 per cent), neither partner expected to marry (70 per

cent), neither partner wanted a child or more children (70 per cent), and the couple

already had children of the union in the household in wave 1 (76 per cent).

Conversely, ongoing cohabitation was relatively less likely among couples in which:

the male partner alone held a degree (for marriage was common amongst this group)

(40 per cent), and one or both partners wanted a child (41‐49 per cent).

As expected, continuing cohabitors had been living together for a longer period of

time than cohabitors who separated or married (means = 5.5 years, 4.4 years and 3.5

years respectively), and the continuing cohabitors tended to be older than those who

separated or married (means = 34.3 years, 32.5 years and 30.9 years respectively).

Overall, these results suggested that couples who married were the most distinct

group in the sense that they differed on a wider range of characteristics than was

apparent for the other groups.

Evidence of screening into marriage unsuitable matches?

It is noteworthy that 34 per cent of marriages occurred after one or both partners

failed to express high relationship satisfaction (n=28). It was predicted that,

compared with other couples who married, these couples would have been older (H

6

)

and would have invested more time in cohabitation (H

7

).

The second of these predictions was supported: by wave 1, couples in which both

partners had expressed high satisfaction had been cohabiting for the shortest period of

time (mean = 2.9 years), followed by couples in which the female alone expressed

high satisfaction (mean = 3.4 years), then couples in which neither partner expressed

high satisfaction (mean = 5.0 years), and finally, couples in which the male partner

alone expressed high satisfaction (mean = 6.6 years). However, these trends are

highly unreliable given the very small number of couples in which partners did not

express mutually high satisfaction.

There was less support for the prediction relating to age. Amongst couples who

married, the female and male partners tended to be older where the male alone

expressed high satisfaction (means = 36.1 and 39.3 years respectively) than in all

other conditions: cases where high satisfaction was expressed by both partners (means

= 29.8 and 32.8 respectively), by the female only (means = 29.3 and 33.6

respectively), and by neither partner (means = 32.1 and 33.3 respectively). Again,

however, the small size of the groups needs to be emphasised.

In short, cohabitations were most likely to convert to marriage when both couples

were highly satisfied with the relationship, and these couples spent the shortest time

living together before they married. Of all cohabitations that converted to marriage,

the very small number in which the male partner alone was highly satisfied had

invested the longest time in cohabitation and the partners tended to be older than other

16

groups who married. However, there was no clear suggestion amongst this small

subgroup that age might be a strong predictor of the “screening in” of unsuitable

matches, where the measure of “suitability” was restricted to relationship satisfaction.

Modeling the determinants of cohabitation transitions

Multinomial logistic regression was used to examine the relative importance of

characteristics of the couples in predicting cohabitation pathways. Given a total

sample of only 398 couples, it was important to pose restrictions on the number of

variables entered in the model (see Long & Freese 2001). The measure of

expectations about marrying (or not marrying) was omitted, given that such

expectations may be somewhat confounded with the pathways taken. In many cases

they would represent a step in, or outcome of, the commitment decision-making

process and the fact that the vast majority who married expected to do so in the

previous wave would seem to offer little of value to the model. Also omitted were the

age differences between partners and employment status of each partner. Neither of

these factors was significant in preliminary runs of the model, and couple-level work

hours were strongly correlated with income.

It seems easier to interpret the probabilities of following the different cohabitation

pathways according to each predictor entered into the model (net of all other factors),

than it is to interpret the multinomial logit model coefficients themselves. Table 2

sets out these probabilities for all except the predictors that were entered as interval

data (female partner’s age and the duration of cohabitation prior to wave 1). The

probabilities of following different pathways according to age and cohabitation

duration are depicted in Figures 2 and 3. The multinomial logit model coefficients for

each predictor are presented in Appendix 2. Unless otherwise specified, all results of

this analysis that are outlined below were significant at the 5 per cent level or better.

As expected, the probability of marriage decreased if one or both partners were

students, but increased when the male partner alone had a degree. However, contrary

to expectations, the likelihood of marriage did not increase when both partners had a

degree. (It was predicted that a degree held by the male partner would increase the

chance of marriage, regardless of the female partner’s educational status.)

While the results for the bivariate analysis (Table 1) suggested that, relative to other

couples, those with high combined annual incomes were likely to marry, and those

with low combined incomes were likely to separate, the multivariate analysis suggests

that the combined incomes affected separation but not marriage (with low incomes

increasing the probability of separating, as predicted). One possible explanation for

these findings is that a male partner’s education washes out the effect of high income

on marriage. Perhaps it is the prospect of a high or at least secure income that is more

important than actual income for marriage. However, low income may create

pressures on relationships that leads to separation. Alternatively, low income may be

linked with other factors (not measured here) that are associated with separation.

Likewise, the bivariate analysis suggested that disproportionate numbers of couples

who agreed that they were in a comfortable or prosperous financial position married.

However, when the other predictors were controlled, such positive appraisals shared

by partners did not significantly affect the probability of marriage, although they

17

reduced the probability of separation. Separation was least likely when both felt

comfortable.

As expected, high relationship satisfaction shared by each partner increased the

probability of marriage and decreased the probability of separation, while the opposite

applied when neither partner expressed high satisfaction (i.e. the less favourable but

mutually shared views decreased the probability of marriage and increased the

probability of separation). Consistent with research suggesting that wives are

considerably more likely than husbands to initiate marital separation, the probability

of separation from cohabitation increased when the male alone expressed high

satisfaction with the relationship. (The increased probability of marriage when the

female partner alone expressed high satisfaction approached significance (p<.10)).

As predicted, the chance of separation increased if only one partner wanted to have a

child or more children. This effect was stronger if it was the male partner who

wanted children. Although the bivariate analysis suggested that couples who shared a

strong desire for a child or more children were most likely to marry, this effect did not

appear to significantly increase the probability of marriage, when the effects of the

other factors were controlled.

On the whole, the effects of family type on cohabitation pathways were consistent

with predictions. Firstly, the probability of marriage decreased if a child had already

been born to the couple by wave 1. In other words, the chances of continuing to

cohabit increased under such circumstances. Secondly, the probability of separation

decreased if the couple had no children in the household in wave 1, but experienced a

pregnancy or birth afterwards. Indeed, the probability of marriage was higher than

that of separation when such a pregnancy or birth took place. It was expected that the

presence of stepchildren in the household would increase the probability of

separation. This effect was in the predicted direction and approached significance

(p<.10).

It was also expected that a previous marriage or cohabitation that did not lead to

marriage would increase the probability of separation on the grounds that such

couples should include an over-representation of people who have difficulty

maintaining intimate relationships. However, neither of these types of previous

relationships significantly affected the likelihood of marriage or separation. Perhaps

the measures used were too crude for this analysis. (With the accumulation of future

waves, the number of couples in which at least one partner has experienced several

former relationships will increase and a more complex measure that takes this into

account the relationship histories of each partner will be possible.)

While the average age of the female partner was lowest for those who married and

highest for those who separated, the impact of the female partner’s age on the

probability of marriage approached significance only (p<.10). Figure 3 suggests a

marginal but diminishing increase in the probability of marriage with age. This non-

linear direction of trends was predicted.

Duration of cohabitation at wave 1 had a significant, non-linear effect on separation:

as expected, the longer the duration, the lower was the probability of separation, with

this effect diminishing as the period of cohabitation increased (as shown in Figure 4).

18

In other words, cohabitants tended to separate early on in the relationship, when

obvious signs of difficulties are likely to surface. Figure 4 also suggests that the

probability of cohabitation continuing to wave 3 increased the longer people had been

living together prior to wave 1. However, it is important to point out that couples who

had cohabited for a relatively long time prior to wave 1 would be a select group ‐ for

the analysis does not include cohabitors who separated or married before wave 1.

Conclusions

Today’s world is changing rapidly and unpredictably on many fronts. Old certainties

have been shaken, and people appear to be treading very cautiously along the pathway

to family formation. Men who see themselves as required to take on the role of

principal and stable sole breadwinner for any future family may be reluctant to enter

into any long-term commitment until they are fairly confident that they can

successfully assume these long-term responsibilities. This now requires considerable

investment in education and career development. The knowledge that they would

probably lose everyday family life with their children should their marriage end in

divorce (a risk that is reasonably high these days) may also encourage men to be

cautious about committing to marriage and starting a family.

Today’s uncertainties are also likely to encourage women to be very careful in any

search for a partner who will be compatible on many fronts, including being a good

provider ‐ and possibly sole provider while the children are very young. Given the

cost of housing and other lifestyle commodities deemed necessary today, women’s

investment in their own education and career are also significant contributions to the

financial wellbeing of any family they have. Should their marriage break down, such

investment may be the only means by which they and their children can avoid poverty

‐ unless a suitable new partner can be found.

Whereas Bumpass and Sweet (2001) argued that cohabiting couples must believe that

marriage will change their lives for them to follow this pathway, Smock et al.’s

(2005) qualitative study of low income (but not poor) cohabitors suggested that

marriage would only occur if and when their financial circumstances had changed.

Likewise, Cherlin (2004) argued that low income couples want to ensure that they are

financially secure before embarking on marriage.

In the present analysis, which covered all income levels, neither combined annual

income nor each partner’s assessment of their financial circumstances had an

independent effect on the probability of marriage, although the chance of separation

increased for those with low combined incomes. Separation may have resulted from

financial pressures (which may have eroded relationship satisfaction) and/or from a

desire to seek a partnership that offers better financial resources. But this is pure

speculation.

On the other hand, the chance of marriage increased when the male partner had a

degree, suggesting that his potential to provide financial security may have been

important. To the extent that an emphasis on the male partner’s earnings reflects an

emphasis on achieving a secure income stream, the results provide some support for

the contention that money matters for marriage. These results are consistent with

those in an Australian study of fertility decision-making, which suggested that both

19

men and women tend to attach high priority the male partner’s job security when

considering the option of having children (Weston, Qu, Parker and Alexander 2004).

(In that study, the most commonly held high priority was to the ability to afford a

child.)

Level of relationship satisfaction was clearly important in predicting both marriage

and separation: when both partners expressed high satisfaction, the probability of

marriage increased and the probability of separation decreased, while the opposite

trends applied when neither partner was highly satisfied with the relationship. Such

findings support the commonly held argument that, nowadays, the quality of the

relationship is of paramount importance in shaping its trajectory (e.g. Giddens 1992;

Clulow 1995; McDonald 1984).

One way of viewing the pathway from cohabitation to marriage is that those who

follow it believe that they have minimised the risks of entering a marriage that is

likely to break down. Furthermore, cohabitations that convert to marriage appear to

share some of the features one might expect of “traditional” marriage. For example,

the finding that marriage was most likely when only the male partner had a degree is

consistent with the traditional pattern of men ‘marrying down’. A well-educated man

is likely to have good career prospects, and thus to be a satisfactory provider. The

relatively low probability of marriage among couples in which both partners have a

degree may be due to the fact that such cohabiting couples delay marriage longer

while the female partner completes a degree or gets established in a career.

Alternatively, it may simply be that more highly educated women are less prone to

marry and see less of a need for the security of marriage.

Where the female partner was satisfied with the relationship (regardless of whether he

was) and where she wanted a child (regardless of whether he did) marriage was

relatively more likely than where the male (at least) was not highly satisfied or where

he was not enthusiastic about having a child. It appears that on these dimensions

(relationship quality and wanting children) the female partner’s views are the key in

the decision to marry.

These patterns are consistent with the interpretation that the move from cohabitation

to marriage reflects traditional gender patterns and represents minimization of risk.

For women at least, marriage is more probable when she is satisfied about the

relationship quality and when she wants a child. A satisfying relationship with a male

who has reasonable income earning prospects is likely to be a less risky context in

which to have children.

Where income is low and there is some discomfort about the couple’s financial

situation, and where at least one partner (especially the female) is not satisfied with

the quality of the relationship, and where there is disagreement about wanting

children (especially where the female partner is not keen to have children) the

cohabiting relationship is more likely to end in separation. This suggests that

marriage (or even ongoing cohabitation) is less likely where there is either economic

hardship or concerns about financial situation. Where she is not satisfied with the

relationship separation is more probable and where she does not especially want

children marriage.

20

Nevertheless, not everyone is cautious about taking risks and some people may prefer

marriage over being single regardless of signs of problems in the relationship. Some

may also believe that marriage may help to rescue their relationship. In fact, some

may believe that they will be able to “change” their partner for the better when they

are married. One-third of couples married despite one or both partners failing to

express high satisfaction with the relationship in the wave prior to the marriage.

These couples had, on average, lived together for a relatively long period before they

married. This may suggest a hesitancy about entering into a marriage when the

relationship is not mutually highly positive and/or a desire to marry after having

invested much time in this relationship.

Little attention has been given in the past to circumstances that might lead couples to

marry even though their relationship during cohabitation has been less than highly

satisfactory. The total number of couples in this analysis was small and the results are

thus unreliable. As the HILDA waves accumulate, the number of such couples will

increase, thereby providing opportunities for understanding processes underlying

“screening in”.

Final word

As noted above, several authors have argued that the stability of modern marriages is

contingent on the quality of the relationship - although financial pressures also appear

to challenge this stability. An emphasis on financial prospects appears to have always

existed. The post-war marriage boom involving a dramatic decline in the age at

which couples married is a case in point: Australia had, at that time, entered a

prosperous period entailing almost full employment. By contrast, the marriage rate at

the start of the 20

th

century was lower than it is today, partly because Australia was

still feeling the effects of the severe economic depression of the 1890s (see de Vaus,

Qu and Weston 2003b).

A number of interacting factors help explain the recent fall in the marriage rate and

increase in age at marriage (see de Vaus et al. 2003b). One of these appears to be the

need to gain qualifications to improve chances of financial wellbeing; another appears

to be emphasis on relationship quality; and a third appears to be a trend towards

cohabitation - which in Australia often seems to be used as a “trial marriage”.

Financial prospects may represent an important factor shaping the perceived quality

of relationship.

21

References

ABS (2001), Census of Population and Housing: Selected Social and Housing

Characteristics Australia, Catalogue No. 2015.0, Australian Bureau of Statistics,

Canberra.

ABS (2002), Marriages and Divorces 2001, Catalogue No. 3310.0, Australian Bureau

of Statistics, Canberra.

ABS (2004a), Marriages, Australia 2003, Catalogue No. 3306.0.55.001, Australian

Bureau of Statistics, Canberra.

ABS (2004b), Births Australia 2003, Catalogue No. 3301.0, Australian Bureau of

Statistics, Canberra.

Baxter, J. (2005), “Mothers’ employment transitions following childbirth”, Family

Matters, No.71, pp.11-17.

Becker, G.S. (1981), A treatise on the family, Harvard University Press, Cambridge.

Becker, G.S., Landes, E.M. & Michael, R.T. (1977), “An economic analysis of

marital instability”, Journal of Political Economy, vol. 85, pp.1141-1187.

Booth, A. & Edwards, J.N. (1992), “Starting over: Why remarriages are more

unstable”, Journal of Family Issues, vol.13, pp. 179-194.

Brown, S. (2000), “Union transitions among cohabitors: The significance of

relationship assessments and expectations”, Journal of Marriage and the Family, Vol

62, pp. 833-846.

Carmichael, G. A. (1995), Consensual partnering in the more developed countries: an

overview, Working Papers in Demography No.56, Division of Demography and

Sociology, Research School of Social Sciences, Australian National University.

Coleman, M., Ganong, L. & Fine, M. (2000), “Reinvestigating remarriage: Another

decade of progress”, Journal of Marriage and the Family, Vol 62, pp. 1288-1307.

Cherlin, A. (2004), “The deinstitutionalization of American marriage”, Journal of

Marriage and Family, Vol 66, pp. 848-861.

Clulow, C. (1995), Women, men and marriage: Talks from the Tavistock Marital

Studies Institute, Sheldon, London.

de Vaus, D. (1997), “Divorce” in D. de Vaus & I. Wolcott (ed.), Australian family

profiles social and demographic patterns , Australian Institute of Family Studies,

Melbourne.

de Vaus, D. (2004), Diversity and change in Australian families: Statistical profiles,

Australian Institute of Family Studies, Melbourne.

22

de Vaus, D., Qu, L. & Weston, R., (2003a), “Premarital cohabitation and subsequent

marital stability”, Family Matters, no. 65, pp. 34-39.

de Vaus, D., Qu, L. & Weston, R., (2003b), “Changing patterns of partnering”,

Family Matters, no. 64, pp. 10-15.

Duvander, A.Z. (2000), “The transition from cohabitation to marriage: A longitudinal

study of the propensity to marry in Sweden in the early 1990s”, Journal of Family

Issues, Vol. 20, pp.698-717.

Ermisch, J. & Francesconi, M. (2000), “Cohabitation in Great Britain: Not for long,

but here to stay. Journal of the Royal Statistical Society (A, 163, Part 2), pp. 153-171.

Giddens, A. (1992), The transformation of intimacy: Sexuality, love and eroticism in

modern societies, Polity Press, Cambridge.

Gray, M., Renda, J., Qu, L., and de Vaus, D. (2003), Changes in the labour force

status of lone and couple Australian mothers, 1983‐2002, Research Paper No.33,

Australian Institute of Family Studies, Melbourne.

Harrison, M. (1986), “Legal process”, in P. McDonald (ed.), Settling up: Property

and income distribution on divorce in Australia, Australian Institute of Family

Studies and Prentice-Hall of Australia, Sydney.

Long, S. & Freese, J. (2001), Regression models for categorical dependent variables

using STATA, Stata Press, Texas.

Lyngstad, T.H. (2004), “The impact of parents’ and spouses’ education on divorce in

Norway”, Demographic Research, Volume 10, Article 5. Online:

http://www.demographic-research.org.volumes/vol10/5/10-5pdf.

McDonald, P. (1984), Can the family survive? Discussion Paper No. 11, Australian

Institute of Family Studies, Melbourne.

Manning, W. & Smock, P. (1995), “Why marry? Race and the transition to marriage

among cohabitors”, Demography, Vol. 32, pp. 509-520.

Oppenheimer, V.K. (1988), “A theory of marriage timing”, American Journal of

Sociology, Vol. 94, pp.563-591.

Santow, G. and Bracher, M. (1994), “Change and continuity in the formation of first

marital unions in Australia”, Population Studies, Vol. 48, No.3, pp. 475-496.

Shehan, C.L., Berardo, F.M., Vera, H. & Carley, S.M. (1991), “Women in age-

discrepant marriages”, Journal of Family Issues, Vol. 12, No. 3 pp.291-305.

Smock, P. & Manning, W. (1997), “Cohabiting partners’ economic circumstances and

marriage”, Demography, Vol. 34, pp. 331-341.

23

Smock, P., Manning, W. & Porter, M. (2005), “’Everything’s there except money’:

How money shapes decisions to marry among cohabitors”, Journal of Marriage and

Family, Vol. 67, pp. 680-697.

Tzeng, M.S. (1992), “The effect of socio-economic heterogamy and changes on

marital dissolution for first marriages”, Journal of Marriage and the Family, Vol. 54,

pp. 609-619.

Watson, N. & Wooden, M. (2002), The Household, Income and Labour Dynamics in

Australia (HILDA) Survey: Wave 1 survey methodology, HILDA Project Technical

Paper Series, No.1/02.

Weston, de Vaus and Qu (2003), Partnership formation and stability, Paper presented

at National Council for Family Relations annual meeting, Vancouver, November 14-

25, 2003.

White, L. (1990), “Determinants of divorce: A review of research in the eighties”,

Journal of Marriage and the Family, Vol. 52, pp. 904-912.

Wolcott, I. & Hughes, J. (2000), Towards understanding the reasons for divorce,

Working Paper No.20, Australian Institute of Family Studies, Melbourne.

Wu, Z. & Balakrishnan, T.R. (1995), “Dissolution of premarital cohabitation in

Canada”, Demography, Vol. 32, pp. 521-532.

24

Figure 1. Pathways since wave 1 (2001)

Figure 2. Cumulative percentage of couples who made transitions since wave 1

Note: Excludes attrition couples.

25

Table 1. Cohabiting couples: Characteristics by pathways from since wave 1

Marriage Separation

No

change Total N

Educational status (%) *

Both had degree or higher 28.8 17.3 53.8 99.9 52

Only she had degree 29.2 18.8 52.1 100.1 48

Only he had degree 46.7 13.3 40.0 100.0 30

Neither had degree 14.2 20.1 65.7 100.0 268

Student (%)

One or both partners studying 20.5 18.7 60.8 100.0 283

Neither studying 20.0 20.0 60.0 100.0 115

Combined annual income (%) *

High (>76k) 26.9 15.7 57.5 100.1 134

Medium (50-76k) 20.8 14.6 64.6 100.0 130

Low (<50k) 13.4 26.9 59.7 100.0 134

Employment (%) *

Both full time employed 35.1 16.6 48.3 100.0 151

Only she full time employed 20.8 25.0 54.2 100.0 24

Only he full time employed 9.9 18.0 72.0 99.9 161

Neither full time employed 11.3 25.8 62.9 100.0 62

Self-assessed financial situation (%)

a

*

Both felt comfortable 29.7 13.4 57.0 100.1 172

Only she felt comfortable 21.2 30.3 48.5 100.0 33

Only he felt comfortable 14.0 27.9 58.1 100.0 43

Neither felt comfortable 12.5 26.0 61.5 100.0 104

Satisfaction with relationship (%)

b

*

Both very satisfied (ratings 9-10) 35.8 10.9 53.3 100.0 137

Only she very satisfied 26.7 17.8 55.6 100.1 45

Only he very satisfied 13.6 27.1 59.3 100.0 59

Neither very satisfied 7.4 27.8 64.8 100.0 108

Expectations of marrying the partner (%) *

Both expected to marry (likely/very likely) 35.4 13.6 51.0 100.0 206

Only one partner expected to marry 4.7 21.9 73.4 100.0 64

Nether 3.9 26.6 69.5 100.0 128

Desire for a child or more children (%) *

Both held strong desire (ratings 9-10) 37.9 12.6 49.4 99.9 87

Only she held strong desire 25.0 27.3 47.7 100.0 44

Only he held strong desire 24.3 35.1 40.5 99.9 37

Neither held strong desire 12.2 17.4 70.4 100.0 230

Family type and recent change (%)

c

*

No children and no change 26.1 17.8 56.1 100.0 180

No children and change (birth/pregnancy) 44.8 3.4 51.7 99.9 29

Intact family and no change 8.5 15.5 76.1 100.1 71

Blended or step family and no change 10.4 31.2 58.4 100.0 77

Other 17.1 19.5 63.4 100.0 41

26

Female partner's age (Mean) 30.9 32.5 34.3 32.5 398 *

Age difference (%)

He 10 or more yrs older 16.3 20.4 63.3 100.0 49

He 5-9 years older 28.0 16.0 56.0 100.0 75

Similar age 19.7 18.4 61.9 100.0 244

He 5 or more yrs younger 13.3 30.0 56.7 100.0 30

Union duration at wave 1 (yrs) (Mean) 3.5 4.0 5.5 4.8 398 *

Marital status

Both never married 23.0 19.4 57.7 100.1 222

One or both ever married 17.0 18.8 64.2 100.0 176

Total

Previous cohabitation not leading to marriage

Both no prior cohabiting relationship 22.5 17.5 60.0 100.0 160

One or both prior cohabiting relationship 18.9 20.2 60.9 100.0 238

Total

N 20.4 19.1 60.6 100.1 398

For continuous variables, ANOVA was used to test whether the means were significantly different.

For categorical variables, chi-square test was applied.

* p <0.0; # P<0.1.

a

11.6 per cent of all couple where either one or both partners did not return the self-completed

questionnaire which contains this question

b

12.3 per cent of all couple where either one or both partners did not return the self-completed

questionnaire which contains this question

c

“Family type” refers to situation at wave 1, while recent change refers to a pregnancy or birth

immediately prior to the wave in which a marriage or separation was first recorded or between waves 2

and 3 if no transition took place.

27

Table 2. Predicted probabilities of each pathway since wave 1

Marriage Separation No change

Educational attainment

Both had degree or higher 15.7 16.7 67.5

Only she had degree + 19.7 16.5 63.8

Only he had degree + 40.2 14.3 45.5

Neither had degree 10.8 13.1 76.1

Student

One or both partners studying 8.3 13.9 77.9

Neither studying 16.7 14.3 69.0

Combined annual income

High (>76k) 15.5 13.7 70.8

Medium (50-76k) 17.1 7.5 75.4

Low (<50k) 9.2 25.8 65.0

Self-assessed financial situation

a

Both felt comfortable 16.8 12.9 70.3

Only she felt comfortable 13.6 28.9 57.5

Only he felt comfortable 9.0 22.4 68.6

Neither felt comfortable 10.2 19.1 70.7

Satisfaction with relationship

b

Both very satisfied (ratings 9-10) 31.4 5.3 63.3

Only she very satisfied 21.5 12.4 66.0

Only he very satisfied 12.4 19.4 68.1

Neither very satisfied 5.3 23.3 71.4

Desire for a child or more children

Both held strong desire (ratings 9-10) 19.6 10.5 70.0

Only she held strong desire 17.8 27.9 54.3

Only he held strong desire 10.2 50.0 39.8

Neither held strong desire 11.1 10.7 78.2

Family type and recent change

c

No children and no change 18.7 13.1 68.2

No children and change (birth/pregnancy) 28.6 1.3 70.2

Intact family and no change 5.7 14.6 79.7

Blended/step family and no change 9.9 28.6 61.5

Marital status

Both never married 13.8 12.7 73.5

One or both ever married 13.6 16.6 69.8

Previous cohabitation not leading to marriage

Neither had other cohabiting relationship 13.6 15.2 71.1

One or both prior cohabiting relationship 13.8 13.7 72.5

Note: The predicted probabilities of taking each pathway are obtained using the estimates of

coefficients in Appendix 2 and the mean values of other predictor variables in Appendix 1.

a

b & c

See corresponding footnotes for Table 1.

28

Figure 3. Predicted probability of each pathway by female partner’s age

Figure 4. Predicted probability of each pathway by length of relationship at wave 1

29

Appendix 1. Descriptive statistics

Mean Std dev

Educational attainment

Both had degree or higher 0.131 0.337

Only she had degree + 0.121 0.326

Only he had degree + 0.075 0.264

Neither had degree 0.673 0.469

Student

One or both partners studying 0.289 0.454

Neither studying 0.711 0.453

Combined annual income

High (>76k) 0.337 0.473

Medium (50-76k) 0.327 0.470

Low (<50k) 0.337 0.473

Self-assessed financial situation

a

Both felt comfortable 0.432 0.495

Only she felt comfortable 0.083 0.276

Only he felt comfortable 0.108 0.311

Neither felt comfortable 0.261 0.440

Not stated 0.116 0.320

Satisfaction with relationship

b

Both very satisfied (ratings 9 & 10) 0.344 0.475

Only she very satisfied 0.113 0.317

Only he very satisfied 0.148 0.356

Neither very satisfied 0.271 0.445

Not stated 0.123 0.329

Desire for a child or more children

Both held strong desire (ratings 9-10) 0.219 0.413

Only she held strong desire 0.111 0.314

Only he held strong desire 0.093 0.291

Neither held strong desire 0.578 0.495

Family type and recent change

c

No children and no change 0.452 0.498

No children and change (birth/pregnancy) 0.073 0.260

Intact family and no change 0.178 0.383

Blended/step family and no change 0.193 0.396

Other 0.103 0.304

Female partner's age (years) 33.3 8.9

Length of relationship at wave 1 (years) 4.8 4.8

Marital status

Both never married 0.558 0.497

One or both ever married 0.442 0.497

Previous cohabitation not leading to marriage

Neither had other cohabiting relationship 0.402 0.490

One or both prior cohabiting relationship 0.598 0.491

a

b & c

See corresponding footnotes for Table 1.

30

Appendix 2. Coefficients of multinomial logit regression of pathways since wave 1

Marriage vs

no change

Separation vs no

change

Marriage vs

separation

Educational attainment

Both had degree or higher 0.491 0.363 0.128

Only she had degree+ 0.773 # 0.409 0.364

Only he had degree+ 1.824 ** 0.598 1.226 #

(Neither had degree)

Either partner studying -0.825 * -0.154 -0.671

Combined annual income

High (>$76000) 0.432 -0.722 # 1.154 *

Medium ($50000-76000) 0.469 -1.378 ** 1.847 **

(Low <$50000)

Self-assessed financial situation

a

Only she felt comfortable -0.011 1.006 * -1.017

Only he felt comfortable -0.596 0.573 -1.169 #

Neither felt comfortable -0.506 0.387 -0.893

Not stated -0.200 -1.454 1.255

(Both felt comfortable)

Satisfaction with relationship

b

Only she very satisfied (ratings 9-10) -0.421 0.814 -1.234 #

Only he very satisfied -1.003 # 1.229 * -2.232 **

Neither very satisfied -1.896 ** 1.361 ** -3.257 **

Not stated -1.978 ** 1.812 # -3.790 **

(Both very satisfied)

Desire for a child or more children

Only she held strong desire (ratings 9-10) 0.161 1.237 * -1.076

Only he held strong desire -0.089 2.129 ** -2.218 **

Neither held strong desire -0.673 -0.087 -0.585

(Both held strong desire)

Family type and recent change

c

No children but change (pregnancy/birth) 0.396 -2.378 * 2.773 *

Intact family, no change -1.338 * -0.051 -1.287 #

Blended/step family, no change -0.532 0.880 # -1.412 #

Other -0.467 0.416 -0.883

(No children and no change)

Female partner's age

Age (years) 0.263 # -0.014 0.277

Age squared -0.004 # 0.000 -0.004

Length of relationship at wave 1

Years -0.058 -0.167 * 0.110

Years squared 0.002 0.007 * -0.004

Either partner ever married 0.035 0.319 -0.284

Either had other cohabiting partner

d

-0.006 -0.128 0.123

Constant -3.892 # -1.194 -2.698

31

Model fit: F( 54, 192) = 2.21

N 398

Reference category is in brackets.

** p <=0.01; * p <=0.05; # p<=0.1

Data have been weighted and statistics test takes into account of the survey design involving

stratification and clustering at wave 1

a

b & c

See corresponding footnotes for Table 1

d

Where cohabitation did not lead to marriage