2. Methodological and data issues
In this section an overview of LSAC is provided and a number of methodological and definitional issues are addressed. LSAC provides Australia's first comprehensive national data on children as they grow up. The study was initiated to examine the effect of Australia's current social, economic and cultural environment on the next generation. More specifically, it seeks to improve understanding of the factors that facilitate or impede healthy early childhood development, to identify opportunities for early intervention, and to inform policy debate more generally.
The data used in this report are from Wave 1 of LSAC, which was conducted in 2004.4 The focus of LSAC is on the early years of children's lives and as such 'the child' is the sampling unit. Two cohorts of children were selected: infants and 4 to 5 year olds. The final Wave 1 sample comprised 5,107 infants and 4,983 4 to 5 year-old children, giving a total sample of 10,090 children.
2.1 Study design, sample selection and representativeness
Although the data used in this report are from only a single wave, and therefore cross-sectional in nature, the study design is longitudinal. The study has adopted a cross-sequential design that follows two cohorts of children whose ages will overlap as the study progresses. These children and their families will be followed at two-yearly intervals until 2010, and possibly beyond.
Children in the infant cohort were born between March 2003 and February 2004, and were aged 3 to 19 months at the interview date. Almost all were aged 4 to 13 months (96.1 per cent, see Table 2.1). Children in the 4–5 year-old cohort were born between March 1999 and February 2000 and were aged 4 years 3 months to 5 years 7 months at the interview date. At the time of interview, the majority were aged 4 years 4 months to 5 years 2 months (97.5 per cent). The sample was selected to be broadly representative of all Australian children in the two age cohorts. The main exception is that children in some remote parts of Australia were excluded because of the high data collection costs in these areas.
The sampling frame for the study was Medicare Australia's database (formerly the Health Insurance Commission). A sample of more than 18,500 children within the LSAC birth dates was selected from Medicare Australia's administrative database.5
Information was collected from the child's parent(s) and from the child (for the 4–5 year-old cohort). Face-to-face interviews were held with the parent who knew the child best (known as the 'primary carer'). Parents were asked to nominate who was the primary carer for this purpose. In 97 per cent of families this was the child's biological mother. In couple-parent families, just 2 per cent of families nominated the father as the primary carer (across both cohorts).6
The primary carer and, where applicable, the primary carer's resident partner (most often the child's other parent) were also asked to fill in separate questionnaires. In addition, the interviewer collected information about the child, their parents, other family members and the neighbourhood. Two diaries ('time-use diaries') were left behind for the parent to record how the child spent his or her time during two 24-hour periods (one weekday and one day on the weekend). Finally, if the parent agreed, a questionnaire was sent to a carer/teacher where the study child spent time in a child care or educational setting.
Calculating response rates is complicated for studies where the sample is drawn from an administrative database. The final Wave 1 sample represents 53 per cent of all families who were sent a letter by Medicare Australia. Refusals were the largest source of sample loss (31 and 35 per cent of infants and 4–5 year olds respectively) followed by 'non-contact'. Non-contact occurred when the address details supplied by Medicare Australia was out of date due to the family moving or when only a post office box address was available and interviewers were unable to locate the respondent. The rates of non-contact were 10 per cent for infants and 14 per cent for 4–5 year olds. The most appropriate response rate is thus constructed by excluding the 'non-contacts'. This results in a response rate of 64 per cent for infants and 57 per cent for 4–5 year olds.7 These response rates refer to interviews with the primary carer.
Infant cohort |
4–5 year-old cohort | |||
|---|---|---|---|---|
| Age in months at Wave 1 interview |
Number of observations |
Age in months at Wave 1 interview |
Number of observations | |
| 3 | 55 | 51 | 57 | |
| 4 | 193 | 52 | 176 | |
| 5 | 322 | 53 | 293 | |
| 6 | 416 | 54 | 414 | |
| 7 | 604 | 55 | 566 | |
| 8 | 798 | 56 | 726 | |
| 9 | 781 | 57 | 733 | |
| 10 | 634 | 58 | 630 | |
| 11 | 504 | 59 | 522 | |
| 12 | 413 | 60 | 421 | |
| 13 | 244 | 61 | 276 | |
| 14 | 94 | 62 | 103 | |
| 15 | 18 | 63 | 21 | |
| 16 | 17 | 64 | 27 | |
| 17 | 9 | 65 | 6 | |
| 18 | 4 | 66 | 7 | |
| 19 | 1 | 67 | 5 | |
| Total | 5,107 | 4,983 | ||
As noted above, parents and carers/teachers were also asked to complete questionnaires. Response rates for these were lower than for the primary carer interviews. The response by the parent(s) to the self-complete questionnaire was highest (85 per cent of primary carers and 79 per cent of the primary carers' partners) and for the time-use diaries (78 per cent), followed by the teacher questionnaire (69 per cent).8
LSAC is broadly representative of the population, with no large differences from ABS Census data on most characteristics. Variables with a close match to Census figures include mother's and father's country of birth and the study child's sex. The main bias in the sample is that the parents of the study child are more highly educated than the general population. The LSAC sample underrepresents children from single-parent families, larger families, those with a non-English-speaking mother, lower income families and those from New South Wales. A detailed discussion of sampling issues and the representativeness of the LSAC data is provided in Soloff, Lawrence and Johnstone (2005). The representativeness of the LSAC data in terms of parental employment rates is analysed in more detail later in this section.
Throughout this report, sample weights have been used to adjust for sampling design and for survey non-response.9 The estimates reported throughout this report take account of the stratification and clustering generated by the survey design.
2.2 Analytical approach and definitions
The LSAC sample was selected to be representative of families and households of infants and 4–5 year olds10 and so it is possible to use LSAC to analyse parental and family characteristics.11 Labour market data are presented separately for mothers and fathers. In addition, for partnered parents, several approaches are used to combine information from each parent (termed 'parental labour supply' or 'family labour supply' in this report). Parental labour supply is also classified according to the primary carer's labour force status.
Throughout this report a number of key variables are used. In this section, the key variables are discussed.
Age of children
Given that employment rates of mothers increase with the age of their youngest child (Gray et al. 2003) it is essential to take into account the age of the youngest child when analysing labour market outcomes. An important issue (related to the design of LSAC) that needs to be addressed when using LSAC to analyse parental labour force status is that nearly half (47 per cent) of the 4–5 year-old cohort live in a family with a younger child. The difference between the age of the study child and the age of the youngest child in the family is shown in Table 2.2. Some 8.8 per cent of 4–5 year olds have a youngest sibling who is aged less than 1 year, 12.8 per cent have a youngest sibling aged 1 year and a further 17.1 per cent have a youngest sibling aged 2 years. While 82.6 per cent of the 4–5 year-old study children were 4 year olds, in only 43.9 per cent of families were they the youngest child. Similarly, 17.4 per cent of the study children in this cohort were aged 5 years, of whom 9.5 per cent were the youngest child. In contrast, virtually the entire infant cohort was the youngest child in the family.12
Infant cohort |
4–5 year-old cohort | |||
|---|---|---|---|---|
| Age of study child |
Age of youngest child |
Age of study child |
Age of youngest child | |
| <1 year | 84.3 | 84.6 | 8.8 | |
| 1 year | 15.7 | 15.4 | 12.8 | |
| 2 years | 17.1 | |||
| 3 years | 7.8 | |||
| 4 years | 82.6 | 43.9 | ||
| 5 years | 17.4 | 9.5 | ||
| Total | 100.0 | 100.0 | 100.0 | 100.0 |
Although the infant sample is restricted to those aged 3 to 19 months, only 20 per cent of the infant cohort was aged 3 to 6 months. The infant cohort also includes some children aged over 1 year of age (7.6 per cent). As such the infant cohort, while representative of all families with an infant within the scope of LSAC, is not representative of families with a 1 year old or under 1 year-old child. Similarly, LSAC is not representative of all families with a child aged 4–5 years, due to the exclusion of children aged less than 4 years and 3 months and children older than 5 years.
Further, combining the cohorts does not result in a sample that is representative of families with a child aged 0–5 years. Families from the 4–5 year-old cohort where the study child was also the youngest child (53.4 per cent of families in this cohort) are representative of families with a youngest child aged 4–5 years (noting the biases mentioned above). Families from the 4–5 year-old cohort where the study child has a younger sibling are representative only of families with a child aged 4–5 years and a younger sibling (and not representative of all families where the youngest child is less than 4 years old).
The approach adopted in this report is to exclude from the analysis children from the 4–5 year-old cohort who have a younger sibling when conducting analysis by the age of the youngest child. This should result in a representative sample of families with a youngest child who is an infant and families with a youngest child aged 4–5 years. In several places, however, the analysis is presented according to the age of the youngest child including families from the 4–5 year-old cohort with a younger sibling. This is particularly the case in the multivariate analyses where age of youngest child can be used as a control variable.
Family characteristics
Throughout the report, the terms 'mother' and 'father' are used to describe the female and male carers of the study child. In the vast majority of cases this was the biological mother (99.5 per cent of female carers) and father (97.9 per cent of male carers) of the study child.13
The face-to-face survey component was conducted with the 'primary carer' of the study child. Single-parent families are defined as those families in which the primary carer of the study child was not living with a partner.14 Couple families are defined to include those who are legally married and those in a de facto or 'marriage like' relationship. The very small number of single-parent fathers in LSAC precludes a separate analysis of this group.
Labour force status
In this report, labour force status is categorised as employed, unemployed and not in the labour force. Much of the analysis combines the unemployed and not in the labour force into the category of not employed. The definition of employed includes those who did any work at all in a job, business or farm, or did any work without pay in a family business in the week prior to the interview. It excludes those who had a job but were on maternity/parental leave or on leave without pay, but it includes those who were absent from their job because of holidays, sickness or any other reason.
Although this definition is similar to the standard ABS definition, there are several differences.15 The differences largely relate to the treatment of those who were absent from their job in the week prior to the interview. The standard ABS definition is that those who had a job but were not at work during the survey week are classified as employed if they meet one of the following conditions: they had been away from work for less than four weeks; or away from work for more than four weeks and had received pay for some or all of the four-week period; or away from work as a standard work or shift arrangement or on strike; or locked out or on workers' compensation and expected to return to their job; or were employers or own-account workers who had a job, business or farm, but were not at work.
For those who had a job but were not at work during the interview week, Wave 1 of LSAC did not collect information on the length of absence from work. Also, except in the case of maternity leave, information was not collected on whether the absence was with or without pay. Finally, for those parents with a job but not at work in the week prior to the interview, there was no attempt to find out about their circumstances in the four weeks prior to the interview in the way the ABS Labour Force Survey does. Consequently, there are a number of ways in which LSAC cannot be used to exactly match the ABS definition of employment and so a number of alternative assumptions were required.
Given the prevalence of women with infants being on leave from work at the time of the interview (see Table 3.1), the decision as to the classification of those on leave has a substantial impact on the employment rates of mothers with an infant. As this report is, in part, about the employment decisions of mothers following the birth of a child, the most useful distinction is whether or not the mother has been in paid employment following the birth of the study child. Consequently, those on paid or unpaid maternity or other types of parental leave, and also those on leave without pay have been classified as being not employed.16 A high proportion of these parents had been absent from work for some time with most not having returned to work since the birth of the study child. In the analysis of leave-taking in Section 3.2, these parents are described as being on a maternity/parental-related leave. Parents who were absent from work for other reasons, or on other types of leave, are classified as being employed on the assumption that these absences (even if unpaid) were only for a relatively short period. Details of the standard ABS questions and the LSAC questions used to measure labour force status are provided in Appendix A.
Given the central importance of labour force status to this report, the employment rates derived from Wave 1 of LSAC are compared to estimates from the 2001 Census and 2004 Labour Force Survey. In making this comparison, it is important to take into account the differences in the definition of employment used in this report and the standard ABS definition. The fact that LSAC is not representative of the youngest infants (those less than 6 months old) should also be noted. Overall, the employment rates of the LSAC parents are comparable to estimates from the 2001 Census and the 2001 and 2004 Labour Force Surveys.17 The conclusion is that the data can be used with confidence to analyse the labour force status of the parents of the study child.