Research report no.13 2005
A snapshot of contemporary attitudes to child support
by Bruce Smyth and Ruth Weston
2. Research design
In December 2004, the Australian Government Department of Family and Community Services (FaCS) commissioned the Australian Institute of Family Studies to conduct an expedited national survey of attitudes to child support in Australia. This study, the Attitudes to Child Support Study, sought to inform the Ministerial Taskforce on Child Support in its review of the functioning of the Child Support Scheme.
The need for fast access to representative, reliable attitudinal data shaped the study's research design. The study had an 11-week time span (from idea to final report) because of the Taskforce's tight timeframe.
The study's temporal parameters required access to an existing pool of suitable attitudinal questions, and 'ready-to-go' samples of (a) separated/divorced parents with at least one dependent child, and (b) adults in the general population. Somewhat fortuitously, the UK Department of Work and Pensions had recently commissioned two related studies of attitudes to child support (Peacey and Rainford 2004; White 2002). These studies provided a set of germane questions on which to draw.
In addition, the Institute had interviewed almost 1,000 separated/divorced parents in September 2003 about patterns of parent-child contact, and most parents had agreed to be recontacted at a later date for future research. This pool of respondents thus formed a readily available sample of separated/divorced parents. A fresh sample of adults in the general population was also obtained. This was a much less onerous task than finding and recruiting a fresh sample of separated parents.
Samples and recruitment
Two samples were drawn. The general population sample comprised 1,001 individuals (55 per cent women; 45 per cent men) aged 18-64 years living in households with telephones. This sampling frame was stratified by gender and geographical location (urban/rural, State/Territory). The second sample comprised 620 separated/divorced parents (54 per cent women, 46 per cent men) aged between 18 and 64 years. These parents had at least one child under 18 years old. 13 Both random stratified samples produced near equal numbers of women and men from all Australian States and Territories, including city and rural areas.
Respondents in the general population sample were randomly selected from the 2004 DtMS Electronic White Pages. This database lists all Australian residential landline telephone numbers. With 14,827 calls, 4,071 private households were identified (as opposed to business or fax numbers, or no answers etc). Of these households, 22 per cent were identified as not containing a person aged 18 to 65 years, while for 49 per cent, the person who answered the telephone refused participation without revealing whether there was a household member of this age.
Response rates can be determined in different ways. If it is assumed that 85 per cent of those households where no information was provided about eligibility were in scope, then the response rate is 36 per cent. 14
Respondents in the separated parent sample were selected through attempts to recontact the 896 (of the original 971) participants in the Institute's 2003 Caring for Children after Parental Separation Project who agreed to be recontacted at a later date. Of these, 678 were contactable, and 92 per cent (n=623) were interviewed, 4 per cent refused, and 4 per cent were away during the survey period.
The original sample was obtained through random digit dialling, and was stratified by gender and geographical location from the population of Australian households with landline telephones. Random digit dialling has a number of benefits over other approaches, including the ability to make contact with unlisted numbers. The proportion of unlisted numbers has increased markedly in recent years, adding bias to samples drawn from the electronic telephone databases.
To obtain the original target sample, more than 163,000 telephone calls were made around Australia, leading to the identification of nearly 70,000 households (43 per cent). Of these households, 77 per cent did not contain a person in scope, while for 15 per cent, the person who answered the telephone refused participation without revealing whether there was a person in scope.
The most favourable response rate (where interviews achieved are calculated as a percentage of interviews plus refusals by a person known to be in scope) is 44 per cent. A more likely estimate, however, is around 26 per cent - assuming that 15 per cent of households where refusal occurred before eligibility could be determined were in scope.15
Data collection
Computer-assisted telephone interviewing (CATI) was used to collect the survey data. CATI offers a number of practical benefits, including substantial cost efficiencies (especially with regard to sampling 'specialised' populations), a relatively fast turn-around from data collection to readiness for analysis, and flexibility in question manipulation and sequencing. CATI also affords excellent data integrity since interviewers are supervised (allowing feedback, support and crosschecks). The CATI software also ensures accurate question sequencing and data entry due to its internal logic checks and verification processes.Interviews were conducted between 19 January and 5 February 2005. The starting date was decided on to avoid conducting interviews in the first two weeks of January when there is a greater potential for temporal sampling bias due to the holiday season (Moser and Kalton 1971).
Respondents in the general population took an average of 12 minutes to complete the interview, while separated parents took an average of 15 minutes (possibly reflecting their greater knowledge and the likelihood that they had much more to say about child support than those who had little or no awareness of, or experience with, the Scheme or the Child Support Agency).
Survey content
The interview schedule had five parts: (a) general perceptions or, in the case of the general population sample, general awareness of the Child Support Scheme; (b) attitudes to child support where fathers were the non-resident parent (c) attitudes to child support where mothers were the non-resident parent; (d) attitudes to broader policy issues; and (e) demographics. (The interview schedules are set out in Appendix A.)
The two schedules were almost identical except for the first section on general perceptions/awareness.16 Respondents in the general population sample were asked whether they had heard of either the Child Support Scheme or the Child Support Agency and, if so, whether they had ever had any contact with the Agency. By contrast, it was assumed that all of the separated parents were aware of the Scheme. They were thus not asked these questions but were offered more direct questions about whether they believed the Scheme was 'working well' and was 'fair'. Both schedules concluded with an open-ended question about whether respondents wanted anything changed in the Scheme. Only respondents in the general population sample who had heard of either the Scheme or Agency were asked this last question.
As noted above, many of the questions for the survey were derived from similar studies conducted recently by the UK Department of Work and Pensions (Peacey and Rainford 2004; White 2002 - see Chapter 1). Both UK studies used face-to-face computer- assisted personal interviews (CAPI) rather than telephone interviews (CATI).
Minor alterations were made to the UK questions to make them more relevant to the family law context in Australia, and to improve their readability for CATI (in which non-verbal cues are not available). One major difference in approach between surveys was that the UK questions focused on situations where the nonresident parent was a father (the most common scenario), whereas some questions in the Australian version included the parallel scenario where the mother was the non-resident parent (see Section 3 of the survey).17 However, time constraints restricted the number of questions that could be gender-reversed.
Analytic strategy
It is important to note that different parts of the analysis are based on different samples. Chapters 3 and 4 of this report focus on two distinct groups from each sample: all women (n=407) and men (n=351) in the general population sample who either were not parents or had not experienced separation from the other parent of their child or children (hereafter called the non-separated sample), and independent groups of resident mothers (n=236) and non-resident fathers (n=149) from the Caring for Children after Parental Separation Project (the separated parent sample). These groups were selected for conceptual clarity, and because they represent the most common groups in each of the respective samples.18 Other groups (for example, separated women and men from the general population, and non-resident mothers and resident fathers) were too small to examine separately.
Chapter 5 explores the responses of wider groups in both samples: respondents in the general population who either (a) had a child under 18 with a parent living elsewhere at the time of interview; (b) had ever paid or received any child support; or (c) had at least heard of the Child Support Scheme or Child Support Agency (n=766); and separated/ divorced parents in the Caring for Children after Parental Separation sample (including the small groups of resident fathers and non-resident mothers) (n=620).
The various samples on which each chapter is based are shown in Figure 2.1.
Demographic profile
Key demographic characteristics of the separated and non-separated groups are presented in Table 2.1. Given that the samples were selected on different criteria, they are likely to differ in important ways - an issue that should be borne in mind when interpreting the results.
Since both samples were stratified by geographical location (State/Territory, city/country), similar patterns of relative proportions in the different localities emerged for each group.
Apart from those aged under 25, there was an even spread of non-separated women and men in the different age groups, whereas the separated respondents were most likely to be aged 35-44 years, followed by 45-54 years. Non-separated women and men were more likely than separated respondents to have a degree.
Not surprisingly, most of those who had never experienced parental separation were married, while most separated parents were single, especially the resident mothers.
While earnings represented the main source of income for most respondents, much the same proportion of resident mothers relied on earnings or government income support. As a consequence of this and their predominantly single status, resident mothers also tended to have lower household incomes than the other groups. (Non-resident fathers in the sample also tended to have lower incomes than the non-separated groups.)
Labour force participation differed along gender lines: the two groups of men tended to be in full-time work; most women were working full-time or part-time, with part-time work being the most common situation for resident mothers - although a higher proportion of non-separated women than resident mothers were not in the labour force.
Methodological issues
Six methodological issues warrant brief mention. First, child support issues are typically complex. However, the survey tool was crude insofar as it mostly offered forced-choice 'yes/no' response options. (The options, 'it depends' and 'can't say', were also allowed for those who volunteered either response. 19 There is little doubt that the complexity of the issues under examination has not been captured by the approach adopted. But a comprehensive study of child support issues was not the focus of this study.
Second, the paradigm on which the survey schedule rests is based on a 'primary care' model (where children are largely in the care of one parent, usually the mother). While there is a broad social and policy push towards a greater sharing of the care of children after separation, shared care is unusual (Smyth, Caruana and Ferro 2004) and could not be reflected in the present framework because of the need to keep the survey short.
Third, attitudes are highly influenced by question wording and order. While every effort was made to minimise order effects, and to ensure that the survey questions were clear, relevant, non-leading, and could be answered by all respondents (especially those in the general population who have little or no awareness of child support issues), it should be borne in mind that slight nuances in question wording and order can lead to marked differences in responses.
Fourth, not everyone is accessible by telephone. The omission of certain groups of people in the population not available through telephone surveys (such as Aboriginal and Torres Strait Islander peoples living in remote areas) sets limits on the generalisations that can be made from the data to the Australian population at large.20 In addition, like other surveys of this type, the response rates were low. Sample biases related to this result are not well understood.
Sixth, the samples of separated men and women in the survey are independent. That is, the men and women had not been married to, or cohabited with, each other. The analysis thus focuses on the views of one parent only - the parent who was interviewed - in examining post-separating parenting arrangements.
- The sample included parents who had never married or lived together. Three respondents were excluded from the present analysis because they were aged 65 or over (the final sample comprised 620 separated parents). Thus both samples had the same upper age limit. It should be noted that by the time of the second interview in 2005, a small proportion of the children had turned 18.
- The 85 per cent estimate is based on the Australian Bureau of Statistics one per cent Census sample file. A far more favourable (but unlikely) estimate of the response rate is based on those known to be eligible. Using this approach, the response rate is 93 per cent (i.e., those interviewed as a percentage of this group and refusers).
- Using data from the Australian Bureau of Statistics, the fieldwork company estimated that 15 per cent of households contacted would 'meet the sample selection criteria'.
- Some demographic questions (eg., gender, age, education) were not asked of respondents in the separated parent sample because this information had already been collected 12 months prior.
- In the UK (Peacey & Rainford 2004), gender differences in sole care were explored with a single question: 'Now think about if the child or children lived with their father rather than their mother. Would your answers on maintenance payments have been the same or different?'.
- There were insufficient numbers to allow comparisons of other relevant groups, such as respondents with stepchildren or new biological children.
- Less than 3 per cent of respondents reported 'can't say' to all except three questions. 'Can't say' responses are only presented in Table 3.1 because they exceeded 10 per cent of all responses. Respondents in the separated sample who articulated a strong need to tell their story were given the option of speaking with one of the researchers at a later date so that they could expand on their responses. These respondents were subsequently contacted and interviewed further.
- Random digit dialling overcomes a number of sampling issues (such as accessing people who have an unlisted number for various reasons including recent relocation, or a desire to avoid harassment, for example, by a former partner or debt collection agencies). The very poor, those with only mobile telephone numbers, and those who have hearing or English language difficulties are among those who tend to be systematically excluded from telephone surveys. These groups warrant separate investigation through more sophisticated recruiting strategies than 'cold calling' by telephone.
