Skip to content

Contents | Download as PDF [1.1MB]

1. Introduction



The Longitudinal Study of Australian Children is the first comprehensive national study examining the lives of Australian children as they grow up. LSAC aims to contribute to a stronger understanding of children's development in Australia's current social, economic and cultural environment. The study was commissioned by and is funded by the Australian Government Department of Families, Housing, Community Services, and Indigenous Affairs. It was announced in the 2000–01 Budget and work commenced on its development shortly after. Wave 1 of the survey was undertaken in 2004 and the data was released for analysis and research in May 2005. LSAC will help governments develop effective policies on early childhood issues, particularly on early intervention and prevention strategies in the areas of health, parenting, family relationships, early childhood education, child care and family support.

LSAC is following the lives of two cohorts of approximately 5,000 children: an infant cohort (0 to 1 years old in Wave 1) and a child cohort (4 to 5 years old in Wave 1). This paper aims to explore how LSAC infants and 4 to 5 year-old children are doing, and to identify factors associated with variations in their functioning. It is based on Wave 1 data, and hence cross-sectional and exploratory in nature. This precludes drawing any implications about causal connections or pathways linking correlates and child outcomes. However, a broad range of factors are examined, ranging from sociodemographic characteristics of families to children's care and educational experiences. A multivariable approach to analysis is adopted, so that unique associations between child outcomes and these factors can be identified. Given the large and representative nature of the two LSAC cohorts, these data provide initial indications about a number of issues of policy relevance, although stronger tests of these must await later waves of LSAC data.

This paper uses the LSAC Outcome Index as the principal indicator of children's current developmental status. The development of the Outcome Index was commissioned by FaHCSIA, to provide an overall summary measure of child functioning (see Sanson, Misson & the LSAC Outcome Index Working Group 2005). The Outcome Index is a composite measure that includes an overall index as well as three separate domain scores, tapping physical development, social and emotional functioning, and learning and cognitive development. The development and characteristics of the Outcome Index are described in detail below, along with discussion of its strengths, limitations, and appropriate and inappropriate uses.

Given that 'outcomes' and their correlates can be more meaningfully assessed at 4 to 5 years than in infancy, the report focuses predominantly on the child (4–5 year-old) cohort, but draws attention to salient findings in the infant cohort also.

back to top

1.1 Report overview

This section includes a brief literature review on existing research on Australian children's development and wellbeing, factors identified in previous research as important contributors to development, and their policy relevance. This is followed by an overview of the design of LSAC Wave 1, and a description of the Outcome Index.

This section is followed by five analytic sections (Sections 2–6). Section 2 describes some of the key demographic characteristics of both the infant and child cohorts, and relationships between these and child outcomes in both cohorts, examining all three Outcome Index domains as well as the overall Index. Nine of these variables are then chosen for multivariable analysis to examine their combined contributions to the overall index.

Section 3 explores the associations of a number of aspects of child care experiences with outcome scores. Section 4 and Section 5 examine the roles of the child's prenatal and postnatal health and exposures, and of maternal physical and mental health on their overall and health outcomes. Section 6 addresses the educational experiences of the child cohort in the home and out-of-home contexts and relates these to overall and learning outcomes.

Section 7 draws these analytic sections together to discuss conclusions and implications of the data.

back to top

1.2 Literature and policy overview

There is extensive international and national attention directed to the development of social policies that focus on 'getting it right' for children during the early years. Children's early experiences are of significant consequence for their later development and learning (McCain & Mustard 1999; McCain, Mustard & Shanker 2007; eds Shonkoff & Phillips 2000). Keating and Hertzman (eds 1999) drew attention to the paradox that, while many post-industrial nations are skilled at creating wealth, there are increasing social and economic disparities resulting in concerns for the health and wellbeing of children who are socially and economically disadvantaged in developed countries. Universal support programs, as well as early identification of children at risk and effective early intervention, can alter developmental outcomes and reduce financial and emotional costs to children, families and communities over time (Williams et al. 2005). Heckman (2000, p. 4) also noted that childhood is a multi-stage process where early investments feed into later investments: '… human capital has fundamental dynamic complementarity features. That is, learning begets learning and skills acquired early on make later learning easier'.

Thus, the early years are an important developmental period that have consequences for children's health and wellbeing across their life course. To make a difference for many outcomes, evidence indicates that the most effective time to intervene to ensure positive long-term outcomes for individuals is in early childhood (McCain, Mustard & Shanker 2007). There is recognition that prevention and early intervention is more cost-effective than later treatment (eds Shonkoff & Phillips 2000). Effective prevention and early intervention are predicated on a good understanding of the complex aetiologies involved in such problem outcomes, as well as an understanding of the pathways to healthy development.

In recent years, governments in Australia have responded to social concerns about children's health and wellbeing. This is reflected in the Australian Government's development of the National Agenda for Early Childhood (Department of Families and Community Services (FACS) 2004b) as well as in the significant investment in the Stronger Families and Communities Strategy (FaCS 2004a). State and territory governments have also sought to improve the coordination and delivery of services to families with young children through a range of policy initiatives that include: Families First in New South Wales; Best Start in Victoria; Putting Families First in Queensland; Our Kids Action Plan in Tasmania; Every Chance for Every Child in South Australia; the ACT Children's Plan in the Australian Capital Territory; and the Vision for Territory Children in the Northern Territory. Besides comprehensive evaluation of existing policies, a strong evidence base developed from research about Australian families and their children is needed to help inform further policy development designed to support young children's development.

Key drivers for LSAC revolve around this need for a strong evidence base, and hence the need to capture information about how children in Australia are faring in the current social, economic and cultural environment, and to understand the factors that impact on their development. As noted in the Australian Institute of Health and Welfare's (AIHW) report, A picture of Australia's children (AIHW 2005), the evidence suggests that most children in Australia are faring well but that there are significant areas of concern. While indices such as infant and child mortality have fallen over the last 20 years, other conditions such as mental health disorders, asthma and allergy, obesity, diabetes, neurological problems such as the cerebral palsies, and learning difficulties have increased or remained at relatively high levels (Stanley, Sanson & McMichael 2002). These problems are not equally distributed across the population but are concentrated in disadvantaged and other vulnerable groups.

To shed light on the multiplicity of influences on children's development, the conceptual model underlying LSAC is a pathways socio-ecological model (Bronfenbrenner 1979; Sanson et al. 2002). As shown in Figure 1, this model situates children's development within the contexts of their family, school and community environments, identifies the interactions that can occur amongst these environments, and points to the influence of broader socioeconomic, structural, cultural and political factors on each context. From each level in the model, the child can be exposed to risk or protection. These influences are understood to occur over time, and to involve both direct and indirect pathways (for example, more distal factors such as parents' work conditions may impact on a child through their influence on more proximal factors such as the parents' relationship with the child or the child's child care experiences). Further, the child is understood to be an active contributor to their own development, with intrinsic characteristics such as genetic endowment and temperamental characteristics shaping their responses to their environments as well as eliciting differential responses from people in these environments (Sanson, Hemphill & Smart 2004).

Figure 1: Socio-ecological contexts shaping children's development

Figure 1: Socio-ecological contexts shaping children’s development

Source: Bronfenbrenner (1979).

Correlates of child development outcomes

Here we provide a brief review of current knowledge about a number of key factors thought to impact on early childhood development, and which are examined in this paper.

Sociodemographic factors

It is clearly important in developing policy and practice to understand how child health and development outcomes are distributed across the Australian population according to background characteristics of the child and family. Adopting an ecological model, these can be categorised as within-child characteristics—such as gender and ethnicity; parental human capital which includes factors like education and occupation; family characteristics such as income and family type (single, two-parent, for example), and neighbourhood characteristics such as liveability and disadvantage which tap aspects of social capital. Previous research indicates that each of these can be related, directly or indirectly, to developmental outcomes, although much remains to be learnt about the processes through which they exert their influence.

For example, international research shows that children from minority ethnic groups tend to have more adverse outcomes than those from the majority culture, although separating out the effects of ethnicity, culture and immigrant status is complex (Quintana et al. 2006). An ethnic background can impact on children's development in various ways, including less exposure to English along with exposure to other languages, differing parental human capital and patterns of child-rearing, lower access to services, and the experience of stigma, disadvantage and discrimination (Helman 2000; Williams et al. 1997). LSAC offers an opportunity to examine both the extent of any developmental disadvantage being experienced by children of non–Australian born parents, as well as potential mediators of these effects.

It is well recognised that Aboriginal and Torres Strait Islander children are faring much worse than non-Indigenous children on most health and wellbeing indicators (ABS 2003a; Silburn et al. 2006). Although the LSAC sampling methodology could not guarantee a representative sample of Indigenous children, the numbers recruited were marginally above population rates (Soloff et al. 2006) (see Section 2), and the rich set of data collected allow some preliminary insights into the factors contributing to outcomes for this group.

Parental human capital has been consistently linked to child outcomes, particularly in the educational arena (for example, Zubrick et al. 1997). Mothers' educational background provides a key resource for children as they develop, and is a critical component of psychological capital. Similarly, the occupational status of parents' work is linked not only to financial security but also to psychological capital, offering a range of experiences and social connections to the growing child. The labour force participation of mothers of young children has been implicated in both positive and negative outcomes, varying by child gender and type of outcome (Hoffman & Youngblade 1999; Sanson et al. 2002).

Due largely to high rates of divorce and separation, there is increasing diversity of family types in Australia and, hence, a need to understand the implications for child development. While children in single-parent and stepfamilies tend to have more adverse outcomes, existing research suggests that the relationship between family type and child outcomes is mediated by factors such as parent–child relationships, social support, family instability and financial stress (Wise 2003). LSAC provides an opportunity to gather national data on the strength of associations between family type and child outcomes, and to identify important intervening variables. Such data can then provide policy guidance to inform family support strategies.

National and international research documents strong negative associations between low income and financial stress and children's development, although theories emphasise different pathways to explain this association (Bradbury 2003; eds Keating & Hertzman 1999). For example, 'investment' theories emphasise deprivation of material resources such as good housing, adequate nutrition, availability of health and support services, and good quality child care; whereas 'parental stress' theories posit indirect paths through parents' psychological and physiological responses to stresses, perceived inequality, limited control over personal circumstances, and social exclusion (Bradbury 2003; Wilkinson 1999). LSAC can, firstly, document the association between income and financial stress and child outcomes and, secondly, explore a number of competing hypotheses to explain such associations. Findings will have relevance to policy options such as directing resources to children and families, or seeking to improve labour market outcomes for parents (Bradbury 2003).

Further, characteristics of the neighbourhood have been shown to be associated with child outcomes, often indirectly through their impact on family functioning and differential access to services (Brooks-Gunn et al. 1997). LSAC includes a number of relevant measures including parents' perceptions of the liveability of their neighbourhood; urban or rural location; remoteness; and the ABS Socio-Economic Indexes for Areas 2001 (ABS 2003b).

There are a number of parent and family factors which have less well-established links to child outcomes, but which are nevertheless important to investigate for their associations in a population study such as LSAC. The number of siblings a child has and the size of their household could similarly have positive or negative effects. Large households can be noisy and stressful, and may entail less parental contact and support for each individual child, but can also provide the child with more social interaction and social support. Analyses of the British National Child Development Study, which has followed a 1958 birth cohort, have found relatively strong family size and sibling number effects on children's cognitive, physical and social development, with effects strongest in the early years (Fogelman 1975; Grawe 2005). Grawe interpreted findings as supporting an economic 'quantitative–qualitative trade-off' model, with effects being due not to financial constraints but the constraints of parental time investment in each child. In contrast, the Dunedin Multidisciplinary Child Development Study (Silva, McGee & Williams 1982) found no associations between family size and birth order and socioeconomic status, height and intellectual development. Overcrowding could make the household more chaotic and stressful and a few studies have found a link to poorer health and other outcomes (Office of the Deputy Prime Minister 2004; Reynolds 2005).

Section 2 reports analyses of the association of outcomes with these sociodemographic factors. The data can speak to important policy questions through identification of their relative contributions, for example: Which factors have unique associations with child outcomes, and which appear to be mediated by other influences? What is the relative contribution of parental human capital and community level factors? Are some domains of development more influenced by sociodemographic variables than others? Can sociodemographic influences be detected in the first year of a child's life, or do they only emerge in the older cohort? Do sociodemographic variables act by increasing the numbers of children with adverse outcomes (that is, in the bottom 15 per cent of the LSAC range) in the context of disadvantage; and/or are children from more advantaged backgrounds more likely to have particularly positive outcomes (that is, in the top 15 per cent of the LSAC range); and/or can the effects of sociodemographic variables be seen across the whole range of outcome scores? As noted above, the reader is cautioned that these cross-sectional data do not support causal interpretations, which must await longitudinal data.

Non-parental child care

The experience of non-parental care is common for many Australian children in the first few years of life. According to figures released by the AIHW (AIHW 2003), 34 per cent of infants under 1 year of age experience regular non-parental care, and this figure rises to 88 per cent for 4 year olds. However, there is much variation in the type of care settings children experience (for example, home-based versus centre-based care), in the amount of time they spend in care, in the stability of their care arrangements—which is reflected in the number of different care arrangements they experience at any one time, and in changes to care over time—and in the quality of the care they receive (in terms of structural aspects of quality, like staff training, group size, and child–staff ratios, or process aspects of quality like the activities provided for children and the responsiveness and emotional quality of staff–child interactions). These different characteristics of child care each affect the experience of children in care and, therefore, must be considered in any analysis of the impact of child care on the health and development of young children. Different characteristics of care are also linked to important policy related issues, including what types of care and what aspects of care quality are most beneficial for children at different ages, whether care in the earliest years of life is detrimental to children's development, and whether long hours of care or unstable care pose risks for children's health and development in the early years.

Results of research from overseas indicate both benefits and risks associated with child care. However, findings across studies are often inconsistent and vary with the specific outcomes investigated. For example, while few studies have investigated children's general physical health in relation to child care, it is generally accepted that attendance at centre-based care settings exposes children to more communicable diseases (like ear, gastrointestinal and upper respiratory tract infections) than attendance at home-based care or care by parents. This association was confirmed by the NICHD Study of Early Child Care and Youth Development (NICHD Early Child Care Research Network 2005a), which reported that 3 year-old to 4 year, 6 month-old children in large group care had higher rates of infection than children in home-based care settings.

Studies of child care in relation to cognitive and language development in young children show more mixed results, with some studies reporting cumulative positive impacts of high quality centre-based care on children's cognitive and language development (Duncan 2003; NICHD Early Child Care Research Network 2000, 2002), and others reporting negligible or non-significant effects (Merrigan & Lefebvre 2002). Studies investigating aspects of children's social–emotional development have produced similarly mixed findings. The NICHD study (NICHD Early Child Care Research Network 1997, 2001) and a smaller-scale Australian study (Harrison & Ungerer 1997) found no negative impact of child care on mother–child attachment relationships in the early years. However, studies of problem behaviours like aggression and non-compliance in children at 4 to 8 years of age have reported more problem behaviours linked to the experience of early, extensive and continuous care (NICHD Early Child Care Research Network 2003).

While the results of overseas research provide an important framework for Australian studies of child care, the results cannot be assumed to generalize directly to the Australian context. Furthermore, in order for the effects of child care on children's health and development to be clearly understood, child care influences must be evaluated within the context of other important influences on children's development that include the child, parent and family characteristics discussed above, as well as family functioning measures such as parenting and family conflict. For example, Burchinal and colleagues (Burchinal et al. 2000) analysed three large North American data sets to investigate the hypothesis that high quality care would be particularly beneficial for children from high risk families (for example, poor, minority ethnic background). They reported that while good quality care enhanced developmental outcomes for all children, the positive effects of quality were stronger for measures of language development for children from primarily African American and Hispanic backgrounds compared to children from white, non-Hispanic backgrounds. Finally, it is important to acknowledge that child, parent and family influences sit within broader social, economic and political contexts that affect parents' working lives and access to child care services, as well as providing a framework of cultural values regarding appropriate parenting and outcomes for children. Research must acknowledge this diversity of influences in order to provide a meaningful assessment of the impact of child care. LSAC represents the first large scale, nationally representative study of child care to address these issues in the Australian context. The analyses reported in Section 3 provide an indication of the concurrent relationships between child care experiences and children's outcomes, taking account of much of this complexity.

Children's health

The right of every child to enjoy the highest attainable standards of health is enshrined in the United Nations Convention on the Rights of the Child (United Nations 1989). Over the last century, there have been remarkable gains in children's physical health—including reductions in infectious diseases, malnutrition, perinatal mortality, and death rates in children experiencing chronic conditions. Nonetheless, physical health remains of considerable concern to parents and to policy makers and some morbidities have risen markedly. Children continue to experience high levels of disability and its effects, with survivors of very preterm births contributing significantly to this pool. Children who live with, rather than die from, chronic conditions such as cancer, cystic fibrosis and diabetes experience lifelong impacts (some of which are only now being delineated) which extend into all aspects of their wellbeing. Like other nations, Australia is in the grip of an obesity epidemic. Although its outcomes are yet to be realised, Australia now has one of the steepest rates of increase in childhood obesity in the developed world (Lobstein et al. 2004). Rates of asthma may have peaked, but diabetes and anaphylaxis are continuing to rise. Social disparities in physical health appear to be widening, and links between physical and psychological health and wellbeing are becoming more evident and their biologic bases more clearly delineated.

Understanding the overall impact of health conditions on children's life experience, and using this and other information to improve outcomes through both prevention and appropriate, effective intervention at the earliest possible stage, are important challenges for population health care systems. With the availability of new measures of health and wellbeing, it is now possible to examine the impact of various physical and mental health conditions. Importantly, LSAC offers an opportunity to explore contextual influences on multiple aspects of health and wellbeing. Factors that are considered to be important contributors to children's overall health include preterm birth and low birth weight; early growth patterns and nutrition, particularly breastfeeding; prevention of disease through full immunisation; good oral health from an early age; illnesses such as asthma, and injuries; and healthy lifestyles, including health nutrition and physical activity at all ages (Victorian Government DoHS 2006). Though not all could be considered in this first cross-sectional wave of data, all will be able to be considered as the study matures. Importantly, use of a single but multi-dimensional measure such as the Outcome Index means that their impacts on children's physical health, social–emotional wellbeing and learning can all be examined simultaneously. This is in keeping with the holistic conceptualisation of children's health as 'a state of complete physical, mental, and social wellbeing and not merely the absence of disease or infirmity' (World Health Organisation 1948).

Mothers' health

The development of life-course epidemiological techniques has focused attention on the intergenerational transmission of both good and ill-health. Parental behavioural choices such as smoking and alcohol intake may directly influence both the physical and mental health of their children (for example, the known direct links between maternal smoking during pregnancy and small-for-gestational-age in infancy, attentional problems in childhood and asthma), which can then be compounded by the increased likelihood of children's own later behavioural choices (that is, to take up similar risky health behaviours as young adults), thus potentiating the problem across multiple generations. This first wave of LSAC offers opportunities to probe existing baseline relationships between such variables, with a view to tracking longer-term relationships as the children mature to a point where they may themselves commence similar risky behaviours.

Much attention has been given to how social circumstances impact on children's development and other outcomes and how maternal factors during pregnancy impact on specific perinatal outcomes such as preterm birth. Perhaps surprisingly, the literature regarding the overall impact of a range of past and current maternal health and lifestyle indicators on children's outcomes has received very little attention, possibly because of the lack in most studies of composite measures such as the Outcome Index. LSAC offers the opportunity to consider intergenerational relationships between mother and child on indicators including pregnancy-related problems, current maternal global health, mental health, habitual fruit and vegetable intake, enjoyment and frequency of physical activity, weight status, and smoking and alcohol intake. A particular strength is the potential to see in which domains health in one generation (the mother) is most strongly related to outcomes in the next (the child).

Children's home learning environments

A significant issue for social policy in relation to children's early learning and development relates to the nature of early educational opportunities available to children before they begin school. LSAC provides a unique source of data to examine the variations in children's home learning environments and how these variations relate to children's development. Important research questions include: What is the nature of family activities and resources that support children's early learning at home? How do children's early learning opportunities in the home environment relate to developmental outcomes? Such analyses that investigate the effects of home learning environments on children's development also need to take account of child characteristics (for example, gender), parental characteristics (for example, maternal education), as well as family sociodemographic characteristics (for example, income) in accounting for variations in children's competence.

Early home learning experiences have been shown to make significant contributions to young children's early learning competence and subsequent educational success (Foster et al. 2005). At the heart of Bronfenbrenner's ecological model (Bronfenbrenner 1989, 1993) are 'proximal processes' which are the everyday interactions between the children and the symbols, objects, and other people in their everyday contexts. Typically occurring activities that involve children and those around them (for example, parent–child interactions around routine activities, reading, watching television) are the 'engines of development' (Bronfenbrenner 1995).

Judge (2005), in an analysis of data from the Early Childhood Longitudinal Study-Kindergarten (ECLS-K) in the United States, found strong links between children's early learning competence and the home literacy environment. The combination of frequent reading and access to books was linked to children's learning competence in reading and mathematics, as well as resilience among children at risk, as measured by sociodemographic factors of mothers' education, welfare support, primary language other than English and family structure. In a large international study of children's early literacy competencies, it was found that children who had higher reading achievement were read to more often at home prior to beginning school, and had many more children's books in their home than children who had lower reading achievement; higher achieving children were also more likely to have access to a computer at home (Mullis et al. 2003). Bus, van Ijzendoorn and Pellegrini (1995) reviewed the literature related to parent–child book reading prior to school, and concluded that such activities contributed to child outcomes such as language growth, emergent literacy, and reading achievement. In this report, the nature of home learning environments, including literacy activities, are examined and related to how children's experiences predict developmental outcomes.

To conclude this review, we reiterate that analyses of Wave 1 are of necessity exploratory and do not allow causal connections to be drawn. By reporting both univariate and multivariable analyses, it is possible to identify factors which appear to have direct associations versus those with indirect (possibly mediated) associations with child outcomes. However, in many cases there may be some ambiguity about the direction and nature of effects. Hence we discuss 'associations' rather than 'influences' or 'causes'. Subsequent waves of LSAC data will support more confident statements about causal directions and hence will provide stronger answers to the sorts of policy questions raised here.

back to top

1.3 LSAC sampling and design overview

Full descriptions of the background, sampling and design of LSAC can be found in Sanson et al. (2002) and Soloff, Lawrence and Johnstone (2005).

In brief, a cross-sequential design was chosen for LSAC, with two cohorts each to be followed over four biennial waves of data collection. The Health Insurance Commission (now Medicare Australia) agreed that the sampling frame for LSAC could be based on the Medicare database, the most comprehensive database of Australia's population. Children in the scope of the survey were infants aged 3 to 19 months (infant cohort) and children aged 4 to 5 years (child cohort). A target sample of 10,000 was sought, equally divided between these two cohorts.

A two-stage clustered design, based on postcodes, was chosen to permit community level effects to be measured and analysed, and to allow cost-effective face-to-face interviewing. Every effort was made to ensure that the sample chosen would be as representative as possible of Australia's infants and 4 to 5 year olds. The first stage of sampling entailed selecting postcodes; the second stage sampled children within these postcodes. Children in both cohorts were selected from the same 311 postcodes. An average of 40 children per postcode in the larger states and 20 children per postcode in the smaller states and territories were selected for the study.

Stratification was used to ensure proportional geographic representation for states/territories and capital city statistical division/rest of state areas. Postcodes were randomly selected with probability proportional to size selection where possible, and with equal probability for small population postcodes. Children were randomly selected with approximately equal chance of selection for each child (about one in 25). Due to difficulties of recruitment and excessive data collection costs, some remote postcodes were excluded from the design, and the population estimates have been adjusted accordingly.

The selection of children and corresponding fieldwork occurred in four phases. This was done to enable sample selection of children born across all months of the calendar year, to attempt to reduce the age range of children at interview, and also because some of the target population had not been born at the time of the first phase selection.

After excluding non-contacts, the achieved response rate for the infants was 64 per cent and 57 per cent for the 4 to 5 year-old children. Broadly, the LSAC sample is representative of the Australian population with no large differences from Australian Bureau of Statistics (ABS) census data on most characteristics. Children with mothers or fathers who had completed Year 12 are a little overrepresented in the final sample. Infants with no siblings are underrepresented (by 3 percentage points), while 4 to 5 year olds in couple families are overrepresented and those in sole parent families underrepresented (by 4 percentage points each). Sample weighting has been used to account for these small differences (see Soloff et al. 2006 for details of weighting procedures, and Soloff, Lawrence and Johnstone (2005) for further details on the design and sample).

Study informants for Wave 1 included:

The primary respondent is the child's primary parent (Parent 1) or main care giver. This person is typically the child's biological mother, but is defined as the person who knows most about the child and their birth, history and current routines.

For the first wave of the study, the base design data collection entailed an interviewer spending one to two hours in the home to:

Full information about the interviews and their content is available in Soloff et al. (2003).

back to top

1.4 the LSAC outcome Index

Before describing the Outcome Index created for LSAC, here we discuss a variety of other composite measures in more general terms to identify how they are derived, the purposes they can serve, and the limitations to their use.

Composite measures and children's developmental outcomes

Besides the ecological model adopted in LSAC to examine influences on the child, the study adopts a holistic view of children's development. It attempts to tap all developmentally-salient aspects of a child's development, from physical health through social and emotional functioning to cognitive development. Given the breadth of policy and scientific interest in the study, and of the research questions it is designed to address, it is important to have relatively fine-grained measures of each of these outcome domains. However, for a broader-brush understanding of how well children are faring and of the important influences on their development, a summary measure of developmental outcomes is also valuable. Nardo et al. (2005) note that a composite indicator provides a report on the 'big picture' and is easier to interpret than trying to find a trend in many separate measures. Composite indicators are designed to raise awareness, but cannot in their own right give the understanding that only individual indicators can reveal. Developing composite indicators of child development is challenging. Measures of child development are not always well understood and the measurement is often indirect, more complex, and less precise.

Most composite indices are based on data aggregated at the national or community level. The best known example may be the United Nations' Human Development Index, which is based on three national indicators: longevity (life expectancy at birth), educational attainment (adult literacy and enrolment ratios in educational institutions), and standard of living (Gross Domestic Product or GDP per capita). A widely used Australian example is the ABS Socio-Economic Indexes for Areas 2001 (ABS 2003b) Index of Relative Socio-Economic Disadvantage, which incorporates census data, aggregated at the Collection District (CD) level or above, on multiple variables including levels of education, income, employment, occupational status, and living conditions.

Since the late 1970s, there has been significant attention to the use of social indicators to monitor children's wellbeing. They are now used in a variety of ways with increasing degrees of sophistication and complexity to inform understanding of the diversity in the lives of children. Social indicators are used to measure the social conditions of children's lives, their health status, and their cognitive and social–emotional competencies. These measures are primarily used to track population changes over time or draw comparisons between children's wellbeing across national contexts. The rationale for their use includes the importance of understanding the social conditions of children's daily lives, and changes in those social conditions over time, to monitor progress towards desired societal goals that relate to wellbeing, and to provide a means to evaluate social policies intended to improve outcomes for children (Land et al. 2007). While indicators can be single measures of social conditions or developmental competence, more consideration is now being directed to developing aggregated or composite indices which combine a number of indicators to provide an holistic picture of children's wellbeing (Ben-Arieh & Goerge 2001). Composite indices can be used descriptively to inform social policy development or relationships can be modelled between social conditions and other family influences that are hypothesised to affect child outcomes.

The use of composite indices of children's wellbeing is primarily used to inform policy development with respect to children and families (Ben-Arieh 2006). They can describe competencies and monitor and track outcomes over time. They can be used to set goals and focus policy directions and activities, as well as providing a means by which policies and programs can be evaluated (Brown & Corbett 2003). Communicating evidence on child outcomes through the use of composite indicators can provide a starting point for discussion about the lives of children. However, composite indices cannot on their own shed light on specific problems which single measures as indicators can reveal. In the development of composite indices, subjective decisions are required in the selection and weighting of individual indicators. Therefore, caution is needed in interpretation of many indices when multiple domains are included in their composition.

Most composite indices of children's wellbeing are based on aggregated statistical information from a variety of measurement sources, and used to measure children's wellbeing within countries. For example, the Child Well-Being Index (CWI) provides a composite index of trends in the wellbeing of America's children (Gilliam 2005). It draws its indicators from a range of national statistical surveys and uses 28 key indicators across seven quality of life domains (family economic wellbeing, health, safety and behavioural wellbeing, educational attainments, community connectedness, social relationships, and emotional and spiritual wellbeing). Similarly, the Bradshaw index of child wellbeing compares the wellbeing of children and youth across eight clusters of indicators for the purpose of making cross national comparisons about the lives of children in countries across the European Union (Bradshaw, Hoelscher & Richardson 2007). Like the CWI, it draws on many sources of data. It uses time series data and comparative surveys of children and young people to compare across countries in eight areas of wellbeing based on 51 indicators. The clusters are children's material situation, housing, health, subjective wellbeing, education, children's relationships, civic participation, and risk and safety. In the Australian context, the AIHW uses a large set of indicators to report on the health, development and wellbeing of Australian children aged 0 to 14 years, most recently in 2005 (AIHW 2005). While such national level indices are very valuable for benchmarking, monitoring trends, and understanding socioeconomic gradients, they are not useful for understanding an individual's current functioning or developmental pathways.

There are also examples of indicators which apply at the level of individuals. For example, most intelligence tests summarise scores on multiple subtests into an intelligence quotient for the individual. Similarly, multiple final school examination results are summarised as a Tertiary Entrance Rank score in several states of Australia. However, the only composite measure known to the authors which aims to capture individual children's behavioural and cognitive functioning is the Vulnerability Index that is constructed from data from the Canadian National Longitudinal Study of Children and Youth (NLSCY) (Brink 2003). This index was created to examine factors associated with childhood vulnerability, to describe the extent of variation across communities and provinces in Canada, and to monitor children's outcomes over time and across studies. It incorporates dichotomised (problem/no problem) scores on a range of measures of motor and social development, vocabulary, mathematics, temperament and behaviour problems. A child is considered vulnerable if one or more cognitive or behavioural outcomes are below these set thresholds. Because Wave 1 of NLSCY covers the age range from infancy to 11 years, different measures are incorporated at different ages. As noted by Willms, from an analytic perspective, information is lost in converting continuous measures into dichotomies (ed. Willms 2002, p. 47); and the imposition of cut-offs between 'problem' and 'no problem' are essentially arbitrary (p. 16). However, Willms argued that the index can still be used for purposes such as monitoring, understanding socioeconomic gradients, and determining developmental pathways. Indeed, it has proven to be a very useful tool for such purposes (see ed. Willms 2002). This index now forms part of Canada's national assessment framework and is used to assess the extent of childhood vulnerability in each province.

Determining how childhood experiences shape developmental outcomes and how policy might reflect indices of wellbeing is at the edge of new knowledge. Much is still to be learned about the substantive issues in the value of composite indices and many methodological questions remain difficult to resolve. There appears to be no single best approach to measuring child outcomes through social indices. Different methods are likely to result in somewhat different answers due to different assumptions underlying the frameworks for construction. Therefore, the development and use of composite indices needs to be understood critically.

It should be noted that, with virtually all composite measures, the components are measured on different scales using different units (in the case of the Human Development Index, years lived, percentage rates of literacy, ratios of enrolments, and GDP) and are combined through standardisation procedures which ensure the distribution of all components has the same shape, allowing a meaningful combination of rankings across components. Thus they produce relative measures, rather than any absolute measure of composite attribute being assessed. An exception is the Canadian Vulnerability Index which uses cut-offs to classify children into problem/no problem on each domain, but, as noted above, its author acknowledges that these cut-offs are essentially arbitrary. A further observation is that the elements in a composite tend to have only modest intercorrelations, since they are intentionally tapping different aspects of the complex attribute they are concerned with. Nevertheless, as the review above indicates, they are proving to be very useful instruments for specified purposes.

The LSAC Outcome Index was commissioned by FaHCSIA to provide an overall measure of how well Australian children were doing. As discussed below, the LSAC Outcome Index is partially modelled on the Vulnerability Index, but attempts were made in its construction to improve upon it by (1) maintaining variables in continuous form wherever possible, hence minimizing the use of arbitrary cut-offs, and (2) incorporating measures of positive as well as problem functioning, recognising that most children are doing well and that it is as important to understand the pathways to better than average functioning as to understand the paths to sub-optimal outcomes (Sanson, Misson & the LSAC Outcome Index Working Group 2005). Further, rather than 'Vulnerability', the more neutral term 'Outcome' was adopted.

Conceptual framework for the LSAC outcome Index

As noted above, the Outcome Index is intended to be a composite measure to reflect how LSAC children are developing. LSAC tracks the development of children across multiple domains, and the Outcome Index provides a means of summarising this complex information. In this context, an outcome is an attribute of the child at a particular point in time. As noted above and in contrast to the Vulnerability Index developed in the Canadian NLSCY, the LSAC Outcome Index wherever possible incorporates both strengths and weaknesses, reflecting the fact that most children have good developmental outcomes. Thus the Outcome Index has the ability to identify groups of children developing poorly and those developing well. Guiding principles in its development were that it should contain all dimensions of developmental interest, and that it should focus on actual developmental status and exclude distal factors (income, family structure, for example) that are predictively related to child outcomes.

A full description of the background and calculation of the Outcome Index is contained in LSAC Technical paper no. 2 (Sanson, Misson & the LSAC Outcome Index Working Group 2005). In brief, three domains are proposed to be the major components of current wellbeing and the future capability to be a successful civic and economic participant: health and physical development; social and emotional functioning; and learning and academic competency. Summary scores for each of these domains are calculated, and they are combined into the overall index. Since each of these domains is roughly equally important for a child's developmental wellbeing, they are equally weighted in the overall index. It is acknowledged that children's development is multidimensional and interactive, and that there are 'fuzzy boundaries' between these domains of functioning. At the same time, a child's development may not be uniform across domains—for example, a child may be doing well in language, but have poorly developed motor skills. It is important to capture this variability.

The broad framework for the LSAC Outcome Index is shown in Figure 2. As can be seen, not all of the components that can be measured for the child cohort (4 to 5 year olds) can be measured for the infant cohort, since these outcomes are not observable at such a young age, and/or are not able to be assessed in LSAC. Hence the meaning of the Outcome Index varies to some extent across cohorts in Wave 1. This needs to be taken into consideration when interpreting the infant cohort, including making comparisons with the child cohort.

The final set of variables selected for inclusion in the Outcome Index is listed in Appendix A. In brief, available measures of physical outcomes in infancy were limited to parent report on two measures of health: an overall rating of the infant's health, and an assessment of whether the infant had greater health care needs than the average infant. Social–emotional outcome was assessed using three scales on the parent report: Short Temperament Scale for Infants (Sanson et al. 1987): which assesses the continuum from withdrawing/shy to approaching/sociable tendencies; Irritability, which assesses the degree to which the infant is calm or volatile, and irritable or not irritable; and Cooperativeness, which assesses the adaptability and amenability of the infant. The learning domain was assessed by parent report on the Communication and Symbolic Behavior Scale (Wetherby & Prizant 2001), which taps the infant's emerging communication skills.

For the child cohort, the physical domain has two subdomains: health, which was tapped by the same two items as for the infants' physical domain, with the addition of Body Mass Index (BMI) calculated from direct measurement of the child's height and weight; and motor, assessed through the parent report PedsQL (Varni, Seid & Rode 1999) physical health subscale, which largely assesses motor coordination but also more general health. There were three subdomains for social–emotional domain assessed by the parent report Strengths and Difficulties Questionnaire (SDQ) (Goodman 1997): Social Competence included the Prosocial subscale (assessing the child's propensity to be considerate and helpful to others) and Peer Problems subscale (assessing the child's ability to form positive relationships with other children); Internalising Problems was tapped by the SDQ Emotional Symptoms subscale (assessing the frequency of child displays of negative emotional states such as nervousness and worry); and Externalising Problems was tapped by the SDQ Hyperactivity subscale (assessing fidgetiness, concentration span and impulsiveness) and the SDQ Conduct subscale (assessing the child's tendency to display problem behaviours such as aggressiveness when interacting with others). The learning domain consisted of four subdomains: Language, assessed from a specially adapted form of the Peabody Picture Vocabulary Test (PPVT-III) (Dunn & Dunn 1997), a measure of receptive language directly administered to children; Literacy, derived from parent and teacher ratings of reading skills, and teachers' rating of writing skills; Numeracy, assessed from teacher ratings of five numeracy skills; and Approach to Learning, tapped by Who Am I?, a direct assessment of the child's ability to perform a range of skills underlying school readiness (ACER 1999).

Full descriptions of the variables in the Outcome Index can be found in the LSAC Wave 1 Data Dictionary ‹http://www.aifs.gov.au/growingup/data/datadictionary.html›.

Calculation of the Outcome Index

The calculation of the Outcome Index for the child cohort involved four stages. In Stage 1 all the outcome variables were standardised, by age where necessary (see Endnotes 1 and 2), and combined into subdomain scores. The process of standardisation is described in the text box below. The second involved standardising the subdomain scores (accounting for missing data where necessary) and combining them in domain scores. The third stage involved standardising the domain scores (again accounting for missing data where necessary), and obtaining cut-offs to identify the top 15 per cent and bottom 15 per cent of the sample for each domain. The final stage calculated the final Outcome Index by calculating a continuous index score from the average of the three domain scores. It is also possible to calculate the number of domains on which a child is in the top and in the bottom 15 per cent of the distribution, and to apply cut-offs to identify the top and bottom 15 per cent of the sample for the overall Outcome Index.

The calculation of the Outcome Index for the infant cohort was a slightly simpler process since there are no subdomains (see Figure 2). This means that Stages 1 and 2 for the child cohort could be merged, so that the outcome measures were standardised (by age, and accounting for missing data where necessary) and combined directly into domain scores. Stages 3 and 4 were identical to those described above.

Box 1: Standardisation

Standardisation of a set of variables, for example, BMI or SDQ, entails making the variables consistent in their mean and variability so that they can be more meaningfully compared or combined. The measure of variability used in the calculation is the standard deviation. A standardised value for a particular variable for each infant or child is calculated by subtracting the mean of the overall sample from the original value and then dividing the difference by the sample standard deviation. The resulting value tells us how many standard deviations it is away from the mean.

An empirical rule applies specifically to data that are approximately distributed according to a 'bell shaped' form, namely that: (1) about 68 per cent of values lie within one standard deviation of the mean, (2) about 95 per cent of observations lie within two standard deviations of the mean, and (3) more than 99 per cent of values lie within three standard deviations of the mean.

The new set of standardised (otherwise called z-score) values have a mean of zero and a standard deviation of 1. A z-score below 1 implies the child is below the average for the total sample, and a score above 1 implies they are above the average.

Figure 2: Conceptual framework for the Outcome Index for the infant and child cohorts, showing domains (in uppercase) and subdomains (in lower case)

Figure 2: Conceptual framework for the Outcome Index for the infant and child cohorts, showing domains (in uppercase) and subdomains (in lower case)

The final suite of variables that comprises the LSAC Outcome Index is:

Details of how missing data were managed are provided in Sanson, Misson and the LSAC Outcome Index Working Group (2005). In brief, a score for a domain was calculated if the respondent had one of the component variables present. However, an overall score was only calculated if the respondent had valid data on all three domains. Using these procedures, overall Outcome Index scores were available on 3,783 of the total sample of 5,107 infants, physical domain scores on 5,106, social–emotional domain scores on 4,314, and learning domain scores on 4,476 infants. In the child cohort, overall Outcome Index scores were available on 4,969 of the total sample of 4,983 children, physical domain scores on 4,982, social–emotional domain scores on 4,969, and learning domain scores on 4,982 children.

Table 1 shows that approximately two-thirds of both infants and children did not score below the negative cut-off on any of the three domains, as might be expected. Around one-quarter in both cohorts were below the negative cut-off on only one domain, and about 7 per cent were below the cut-off on two domains. Only 0.8 per cent of infants and 1.7 per cent of children showed pervasive developmental difficulties, being below the cut-off on all three domains. These data indicate relatively low levels of cross over from problems in one domain to problems in another among children in these two age groups. Interestingly, a similar finding emerged using the Vulnerability Index in the NLSCY: only 3 per cent of children were below the cut-offs for both cognitive and behavioural problems (ed. Willms 2002). A similar pattern is apparent in relation to the positive cut-off, except fewer infants were above the cut-off on two domains (as noted above, it was not possible to define a positive cut-off in the physical domain for infants, and hence they could not be above the cut-off on three domains). Overall, these findings indicate that development does not occur uniformly across all domains at these ages, and they illustrate the need to make judgements about a child's overall development on the basis of information on all areas of their development, not a limited set of domains.

Table 1: Percentage of infants and children in bottom and top 15 per cent of scores across domains
Characteristic

Infants

Children

n % n %
Number of domains on which infant/child scored below negative cut-off (that is, lowest 15%) 3,783   4,969  

None

  63.5   66.2

One

  28.7   24.5

Two

  7.0   7.6

Three

  0.8   1.7
Number of domains on which infant/child scored above positive cut-off (that is, highest 15%) 3,783   4,969  

None

  73.3   63.8

One

  24.1   28.0

Two

  2.6   7.5

Three

  –   0.7

Outcome Index limitations

The following characteristics of the Outcome Index derived from Wave 1 data have important implications for its use and interpretation:

More details on the conceptual basis and calculation of the LSAC Outcome Index can be found in LSAC Technical Paper No. 2 (Sanson, Misson & the LSAC Outcome Index Working Group 2005).

Table 2: Skewness of continuous score distributions for the Outcome Index
Score Infant cohort Child cohort
Physical domain –2.24 –1.66
Social–emotional domain –0.23 –0.79
Learning domain 0.11 –0.14
Full Outcome Index –0.48 –0.85
back to top

1.5 Weighting

The analyses in this paper use sample-weighted data. The sample weights in LSAC allow for unequal probabilities of selection into the sample, and for non-response (to account for the known underrepresentation of female carers who did not speak English and/or had not completed high school in LSAC). Statistical methods were used to obtain estimates of standard error taking account of the correlation of responses within postcodes. More details on the LSAC weighting process can be found in LSAC Technical paper no. 3 (Soloff et al. 2006).

Return to top

2. Outcomes for children in differing circumstances

Executive summary