Belinda Hewitt and Janeen Baxter on Understanding family change: life course and longitudinal approaches with quantitative data

Janeen and Belinda discuss the strengths and weaknesses of longitudinal data for examining family change. This is an article from the November edition of Nexus, the newsletter of the Australian Sociological Association, edited on behalf of the Families, Relationships and Gender thematic group of TASA. Understanding family change: life course and longitudinal approaches with quantitative data
By Belinda Hewitt and Janeen Baxter

The formation and dissolution of relationships and families have undergone major shifts over the last several decades. Rates of marriage have declined; the number of couples living together before, or instead of, marriage has increased dramatically; the number of children being born in defacto unions rather than marital unions is increasing; women are delaying child bearing, couples are having smaller families; and rates of marital dissolution have increased (de Vaus 2004). While many approaches have been applied to examining family change, in this essay we discuss the strengths and weaknesses of longitudinal data for examining family change.

Change at the individual level: The value of longitudinal data
A significant development in the field of family research in Australia has been the increased availability of large panel data sets that collect a wealth of information about people’s relationship and family life course transitions as well as other social, health, demographic and attitudinal characteristics (e.g. Longitudinal Studies of Australian Youth, The Negotiating the Life Course Survey, The Households, Income and Labour Dynamics in Australia Survey, The Longitudinal Survey of Australia Children, Longitudinal Study of Women’s Health). There are substantial advantages to using longitudinal data and analytic techniques over cross sectional analyses when examining transitions and pathways. The collection of successive waves of information from the same individuals enables measurement of the temporal ordering and sequencing of events. Thus we can observe people as they transition from one relationship state to another, and as they experience multiple transitions we are able to build a picture of their relationship and family formation pathways. It also enables a better examination of cause and effect, since the causes of behaviours or outcomes must come prior to the effects. Finally, longitudinal data enables the use of advanced statistical methods that control for heterogeneity and unmeasured bias amongst individuals.
Until relatively recently Australian sociologists trying to understand the processes preceding and following life course transitions such as moving into cohabitation, getting married, getting divorced, becoming widowed, repartnering or having children have faced two main problems (Baxter and De Vaus 2005). The first problem has been with identifying the sequence of events. When in the life course did the event occur? And what happened before or after the event? This can be done to a small extent with retrospective data, but retrospective data is subject to recall bias and memory failure, particularly if the event occurred a long time in the past. The second is working out whether selection or causal factors contribute to the observed differences between groups. Selection implies that it is unmeasured or unobserved characteristics of people that underpin differences between groups, rather than being a member of a specific group. Whereas causation implies that an event or a transition resulted in a certain outcome. To further illustrate these points and how longitudinal data can help overcome them we present two case studies using examples from our recent research.

Study 1: Pathways into marriage: the implications for unpaid domestic labour
Understanding the relationship between union type and relationship outcomes is important because we know that there is much greater variation in patterns of union formation and dissolution than in the past, and in particular, a marked increase in rates of cohabitation prior to marriage. As cohabitation prior to marriage becomes more commonplace, what are the implications of different pathways into marriage for relationship experiences, such as the domestic division of labour? Does spending time in a cohabiting relationship prior to marriage lead to more egalitarian housework arrangements after marriage? Previous cross sectional research has found that housework patterns within cohabiting relationships are more egalitarian than in marital relationships (Baxter 2005). But do these patterns remain when couples move on to marriage? Using the first 3 waves of data from the Household Income and Labor Dynamics in Australia Survey (HILDA) we have investigated this question.
In Figure 1 we present the results of analyses examining the impact of relationship transitions into marriage on housework hours. The graphs show some interesting trends:
• Irrespective of relationship status women spend more time doing housework than men.
• Transitions into relationships have a greater impact on women’s time in housework than men’s.
• Adult children living at home do very little housework.
• When men previously living on their own move into either cohabiting or marital relationships they do less housework than when they were on their own, whereas women who make this transition do more.
• Married women and men end up doing approximately the same amount of housework irrespective of whether they cohabited before marriage or not.
Figure 1 Here
Overall, these results indicate important differences in time spent on domestic work amongst single people depending on whether they are living alone or at home, as well as differences between people in cohabiting and marital relationships. In response to our initial question of whether living together before marriage results in more egalitarian housework patterns after marriage, the short answer is “no”. But there is some variation depending on on whether a person has been living alone for some time or whether they have been living at home with their parents prior to forming a union. Those who transition from living at home with parents into a relationship do less housework the year after they get married or cohabit than those who have been living alone. This suggests another possible explanation: hours of housework may change in response to life course pathways as well as how long a person has been in a certain relationship state. These are questions to be addressed in further research as more years of prospective longitudinal data become available.

Study 2: Marital loss: the implications for mental health
Marital status is arguably one of the most important social determinants of health. Over many decades researchers have consistently found that married people have lower levels of depression, anxiety, psychological distress and better overall mental health than the unmarried. Among the unmarried, people who are separated, divorced or widowed tend to have worse mental health than their never married counterparts suggesting something particularly consequential about the loss of a marriage for mental health.
One of the main problems facing researchers trying to understand the impact of marital loss on mental health is whether or not poor health conditions existed prior to the marital loss which then results in the observed differences, or whether the loss of the marriage caused poorer mental health (see de Vaus 2002; Johnson and Wu 2002 for more detailed explanations of these mechanisms). This is usually referred to as a selection problem. In a study using 6 waves of data from HILDA we examine this issue. Longitudinal data enable us to better differentiate between these selection and causal mechanisms because a person’s mental health (and other characteristics) before they experience the transition can be taken into account when estimating the impact of the transition on mental health. The differences that are due to pre-existing conditions and causal mechanisms associated with marital loss can be isolated.
Figure 2 shows the mean level of mental health for people who experience transitions from marriage into separation, divorce or widowhood. These graphs illustrate two different pathways of transitions out of marriage for men and women. The dark line shows a pathway from marriage to separation to divorce over the 6 waves of data; the light line shows a pathway from marriage to widowhood and then remaining widowed. The key points from these graphs are:
• Taking into account levels of prior mental health, marital loss either through separation or widowhood is associated with a sharp decline in mean levels of mental health.
• The decline in mental health is larger for women than men and is also larger for widowed compared to separated persons.
• The year after becoming widowed mental health levels return to levels similar to those prior to the death of their spouse
• The year after the separation mental health levels return to levels similar to those prior to separation,
• The transition to divorce from separation also has implications for mental health. Men’s level of mental health declines and women’s improves.
Overall, these results suggest that marital loss does cause a decline in mental health. This suggests a crisis hypothesis of marital loss on mental health (Johnson and Wu 2002), where once the crisis has passed mental health returns to levels prior to separation or widowhood.
The additional insights obtained from this longitudinal data are that, taking into account mental health and other characteristics prior to separation and widowhood, marital loss has an immediate and significant impact on mental health, but over time mental health recovers.
Figure 2 here

Conclusions
While this essay has been important for highlighting some of the benefits of using longitudinal data, there are still limitations to this approach. Longitudinal studies have a number of weaknesses that may impact on the results and insights that they offer. One of the main problems is the loss of people to follow-up over waves or sample attrition. To compound the issue, sample attrition is often disproportionately distributed amongst those who undergo particular transitions. For example, people who become separated or divorced are more likely to be lost to follow up than other groups, partly because one or both spouses move residence. Another issue is that often relatively small numbers of people experience the transitions of interest. Even in large nationally representative samples with thousands of participants, only a small percent can be expected to have a first birth, or to become separated or divorced each year. This has implications for the statistical power of the models, and makes is more difficult to detect significant effects.
While longitudinal studies have their problems, they are nevertheless one of the main tools available for us to develop our understanding of the factors that contribute to family transitions and also the consequences of life course events for individuals and families. Observing people at only one time point offers a snap shot of a respondent’s position at that moment, but provides less insight into the mechanisms that contribute to differences between groups. Longitudinal data can be of particular importance to sociological study, a discipline where one of the central concerns is to understand social change. The increasing availability of longitudinal data with a wealth of measures of interest to sociologists greatly increases our capacity to understand the causes and consequences of social change and its effects on individuals.


References
Baxter, Janeen. 2005. "To marry or not to marry: marital status and the household division of labor." Journal of Family Issues 26:300 - 321.

Baxter, Janeen and D. A. De Vaus. 2005. "Editors' introduction to the special issue: 'Life pathways: Insights from longitudinal research'." Journal of Sociology 41:338-342.

Baxter, Janeen, Michele Haynes, and Belinda Hewitt. forthcoming. "Pathways into Marriage: Cohabitation and the Domestic Division of Labor." Journal of Family Issues.

de Vaus, D. A. 2002. "Marriage and mental health." Family Matters 62:26-32.

—. 2004. Diversity and change in Australian families: Statistical profiles. Melbourne: AIFS.

Hewitt, Belinda, Gavin Turrell, and Katrina Giskes. 2009. "Marital loss and mental health: Gender and the importance of social support." in HILDA Survey Research Conference 2009. The Univeristy of Melbourne 16th and 17th July 2009.

Johnson, David R. and Jian Wu. 2002. "An empirical test of crisis, social selection and role explanations of the relationship between marital disruption and psychological distress: A pooled time-series analysis of four-wave panel data." Journal of Marriage and Family 64:211-224.