Researching family stability

The Australian Institute of Family Studies has as its principal task the conduct of research into factors contributing to family stability. In its 1999 Research Plan, the Institute defines a stable family as one which (a) supports the wellbeing of its individual members (especially children), (b) stays together for a sustained period of time, and (c) contributes positively to the wider society of which it forms a part (the Research Plan can be consulted on our web site at http://www.aifs.gov.au/).

These three aspects of family stability are addressed in the three research programs which have now been established at the Institute:

• The Children and Parenting program asks what types of family and child care arrangements are most likely to maximise the wellbeing of children.
• The Family and Marriage program asks what factors are most likely to promote happy and stable family unions which persist through time.
• The Family and Society program investigates the conditions under which families can best contribute to good citizenship and a vibrant community life.

The principal research tool by means of which these three programs will address their core questions is a new panel survey: the Australian Family Panel Survey. It is intended that this survey should enable us each year to monitor and analyse changes in the quality of family life as well as tracking shifts in Australian family values.

The practical advantages of a panel survey

Since it was established in 1981, the Australian Institute of Family Studies has conducted a number of large scale surveys. In most cases, a new sample of Australian families was selected each time. These people were interviewed, their answers were analysed, the results were written up, and that was the end of the project. We may call this a ‘flat’ or one-dimensional survey design, for each project captured information on one issue from one sample of people at just one point in time.

The Institute plans to replace these one-off surveys with a single panel survey. People will still be selected randomly for inclusion, but they will then form a ‘panel’ of respondents, and we shall return to this same panel each year. As people drop out of the panel (e.g. they emigrate, or die, or decide they no longer wish to participate), so they will be replaced by new recruits who (as far as possible) match them in terms of their social characteristics.

There are a number of practical advantages in this move to a single panel:

• Economy of effort: Most surveys ask a common set of questions (e.g. age, occupation, family structure). The new panel means that we shall no longer have to ask the same questions over and over in each different project. It also means that we should achieve low refusal rates, for all participants will already have agreed that we can recontact them, and the recurring interview load also makes for much greater predictability in the planning of the annual use of the Institute’s computer-aided telephone interviewing system (CATI).
• Swift analysis to inform current issues: The Institute is often asked to provide specific, survey-based information relating to some contemporary matter. The development of the panel does not guarantee that we shall in future have such information at our finger tips, but it should make it possible for us to call up a relevant sub-sample of the panel to generate the specific data that are required much more quickly than in the past.
• Ready identification of specific sub-samples: The panel will also make it easier to identify samples representing a particular sub-group of the population such as retired people or single parents. Generating such samples from scratch involves many wasted phone calls, for these people are scattered relatively thinly in the population, but in future we shall know in advance which people to call.

The major advantage of moving to an annual panel survey, however, is not so much practical as analytical, for as time goes on, the panel design will enable us to accumulate information stretching back over a period of years, and this will enormously enhance the quality of the analysis which we can achieve.

The analytical advantages of a panel survey

One of the problems with one-off surveys is that it is difficult to analyse change. Retrospective data are notoriously unreliable, for respondents’ recall of past events can be very shaky, and it is impossible to get valid and reliable information on things like people’s past attitudes. In a panel survey, however, things get recorded as they occur, and as time passes, it is possible to trace patterns of change and to see how events at time A may have contributed to outcomes at time B.

The advantages of this for the kinds of projects the Institute needs to carry out are obvious. In the Children and Parenting program, for example, it will enable us to track the impact on children of different patterns and styles of parenting and out of home care. In the Family and Marriage program, it will allow us to test predictive models of marital stability and thus to evaluate the likely success or failure of interventions such as marriage education and counselling. In the Family and Community program, it will allow us to gauge the impact of social policy changes on family self-reliance and on the quality of community life. And so on.

What questions will be asked: Specific, derived and core items

Every year, the Australian Family Panel Survey will consist of three sections:

• Specific items will vary at each sweep. They will be ‘bolted onto’ the core survey each year to reflect the particular data requirement needs of projects running at that time, and they will be non-cumulative.
• Derived items will be integrated into the data set from other sources. For example, each family’s area of residence will be classified by socio-economic status according to the Australian Bureau of Statistics SEIFA codes, and when panel members have children attending pre-school institutions, the quality of this child care will be classified according to the Commonwealth Childcare Advisory Council’s ‘Quality Improvement and Accreditation System’. Information from sources like these will be recorded and logged along with information collected each year from the core questions.
• Core items will be repeated every year, and it is this part of the survey which will therefore provide the key measures when it comes to analysing change over time.

What goes into the core?

One of the lessons of panel studies is that questions are better organised around themes than specific topics. Specific issues can lose their relevance over time and initial measures thus need to be flexible enough to accommodate future developments which cannot currently be anticipated.

In addition to collecting and updating basic socio-economic and demographic information, the core of the Australian Family Panel Survey will revolve around two main themes:

• Quality of family life: There will be some basic and simplemeasures relating to parenting and child wellbeing, family functioning, and family participation in the wider community (the three key concerns identified in the 1999 Research Plan). This will enable us to gauge whether child wellbeing is improving over time, whether family life is strengthening or weakening, and what is happening to levels of civic responsibility. All these measures will need to be simple and readily understandable, and they should be based on indicators which are widely known and respected in the research community.
• Family values: The core will also include a set of attitude questions relating to various aspects of family life in Australia, the way it is changing, and the main influences which impact upon it. The purpose of this is to enable us to track changes in family values over time.

The data collected in the core survey will thus allow us to monitor changes in the quality of family life and changes in family values on an annual basis.

What about the economics of family life?

The survey cannot and should not try to study everything relevant to families. The core focus of the survey explicitly reflects the domain concerns of the Institute’s research as identified in the 1999 Research Plan, and as such, its strength will lie primarily in the analysis of psychological and sociological measures of family stability and wellbeing. Put another way, the principal focus will be on collecting and refining non-economic measures relevant to the quality of family life and family values over time.

While including some basic economic measures, the survey will not attempt to collect detailed information on things like family income and individual work histories. Such information is certainly needed, but the Commonwealth government announced in its May budget statement the establishment of a new longitudinal survey (to be run by the Australian Bureau of Statistics) which will explicitly cover labour market and education participation of families, together with data on family incomes. Given the Institute’s focus on more psychological and sociological measures of family strength and wellbeing, the proposed ABS longitudinal survey should neatly complement the output from the Australian Family Panel Survey without overlapping it.

How big will the panel be, and how will it be recruited?

We shall recruit a panel of up to five thousand families drawn from right across Australia. Because this is a family panel survey, our unit of analysis is in a sense the family rather than the individual. This means that we cannot afford to rely on interviewing just one person from each family, and that in at least some of our panel families, we shall need to seek information from a second adult member and perhaps from older children too. This will enable us to focus on family relationships and family dynamics and to measure things like family functioning from a variety of different perspectives. This research design is quite distinctive among existing panel surveys in Australia, including the new ABS longitudinal survey.

Although the sample will be selected randomly from households across Australia, it will be ‘stratified’ to ensure that certain groups are properly represented. We know, for example, that participation rates of people from lower socio-economic groups tend in all surveys to be somewhat lower than those from higher ones, and that their drop-out rates tend to be higher than average, so groups like these will be deliberately over-sampled. We shall also need to ensure that there are the right proportions of men and women, and people in different age groups, in the final panel.

Once the panel has been selected, the plan is to interview members using the Institute’s Computer Assisted Telephone Interviewing facility (known as CATI). Most panel members will be interviewed once or twice in any one year, but some sub-sets of the panel may be contacted more frequently. Nobody, however, should be contacted more than three or four times a year, for it is important not to abuse the goodwill of those who agree to participate

What will the survey be used for?

The new panel survey is intended to meet two main objectives:

Monitoring family trends and attitudes:

The Institute already monitors Australian family trends (such as fertility rates, marriage rates and divorce rates) through analysis of secondary data sources, and the results of this monitoring exercise will soon be appearing regularly on our web site as well as in an updated and revised edition of the highly successful AIFS publication, Australian Family Profiles (1997).

The new panel survey will enable us to complement this analysis of secondary data with primary analysis relating to measures of child wellbeing, family functioning and civic participation, as well as reporting what families are thinking and saying about the changes affecting their lives. This means that we can each year make available data which not only measure changes in things like marriage and divorce rates, but which also give us insights into the changing quality of Australian family life as reported by families themselves.

As currently envisaged, this new material would be published annually in a report on Australian Family Life and Attitudes which would provide the authoritative guide to how family life is changing and what families themselves think about it. The aim is that this report should be simply written, attractively presented, and made accessible to as wide an audience as possible.

Research on the stability of Australian family life.

Trends and attitudes monitoring is essentially descriptive — it can tell us whether and how things are changing over time. In addition, however, the new panel survey will also represent the Institute’s major vehicle for conducting analytical research — i.e. for answering questions about why things are as they are. In future, most of our projects which require primary survey material to be collected will use this one ongoing panel (unless, of course, a specific research design makes a random sample of Australian adults inappropriate). Over time, this means that we shall not only build up a longitudinal panel which should enable us to answer questions about social causation which flat survey designs cannot address, but also that we shall develop more and more information about our panel as each new project is added to the stock of the ones that went before. Within a few years, this should become an extremely rich and valuable data set which can be made available to the research community at large.

The next steps

The first sweep of interviews for the new Australian Family Panel Survey should begin late in 1999. A lot remains to be done before this, however, and as part of our preparation, the Institute will be consulting widely with potential users and other interested parties about what should go into the core part of the survey and the uses to which the data should be put. Among the issues still to be resolved are:

• What is the sampling unit?: Can we save time and money by interviewing just one person in some of our panel families, or should we always be interviewing at least two family members? Is it better to include more families by cutting down on the number of interviews in each, or to include partners and children in every family in the panel, thereby reducing the number of different families which can be included? How big does the sample of families need to be?
• Overcoming panel attrition: How can panel loyalty best be maintained? Panel attrition is a major problem in longitudinal research designs, so how can it most effectively be limited without exorbitant expenditure?
• Intensity of use of the panel: How many times can panel members reasonably be interviewed in any one year without risking over-exposure? Should interviewing be staggered across the year or is it best all completed in one long call? Is it sensible to repeat all the core items every year, or would it be better to rotate them?
• Use of other data sets: Should some attempt be made to include among the ‘derived items’ built into this data base information held on other data bases linked to individuals? For example, if panel members gave permission, and if proper safeguards were guaranteed, would it be useful, or even possible, to link the survey to Medicare or social security data bases?
• Stratification of the sample: Should the eventual sample include, say, single person households, or should it be limited to family-based households? Should it focus on particular cohorts (e.g. young families, families with adolescent children, families nearing retirement) or comprise a random sample across the range of different family types and life cycle stages? Should some special effort be made to recruit Aboriginal and Torres Strait Islander families to the panel?
• Selection of indicators: What are the most appropriate measures of child wellbeing, successful family functioning and community vitality (the three core themes)? How can we ensure both continuity of these measures through time, and the flexibility to adapt them to changing circumstances? How do these indicators need to be measured in order to enable best use of time series statistics?
• Final decisions on content: Are there important omissions which need to be rectified? Is there any overlap with other surveys or data sources which could be cut out? Are there other uses for a data set like this which should be taken into account before the base questionnaire is finalised?