Introduction

Globally, research indicates that participation in quality early childhood education and care (ECEC) positively impacts child outcomes and thus plays an important role in children’s development (Raikes et al., 2023; von Suchodoletz et al., 2023). This may be particularly true for children with disadvantaged backgrounds in Australia (Lam et al., 2024). Every dollar spent on early childhood education in Australia generates a $2 benefit to the children, their primary caregivers, the government, and business (The Front Project, 2019). However, the interplay between child and family characteristics, attendance patterns, ECEC program quality and type, and their differential impacts on developmental outcomes is complex (Gilley et al., 2015; Ishimine & Tayler, 2014). Research estimating the causal impact of preschool programs reports mixed results. These range from supporting child development in the long term to having no significant – or even adverse – effects on child outcomes (Burchinal et al., 2024; van Huizen & Plantenga, 2018) and further research is necessary, particularly to identify what promotes optimal early learning.

Australia’s early childhood sector is regulated by the National Quality Framework (NQF) (ACECQA, 2023a). The NQF was the outcome of an agreement entered into by the governments of all states and territories in Australia aimed at supporting improved learning and development outcomes for all children within a nationally consistent regulatory system. The NQF introduced national learning frameworks, including Belonging, Being and Becoming: The Early Years Learning Framework for Australia (DEEWR, 2009).

In light of the mandated use of the EYLF (AGDE, 2022; DEEWR, 2009) for the last 15 years, high overall rates of child enrolment in ECEC programs across the country (Lam et al., 2024), and national data from the Australian Early Development Census (AEDC), in this study, we aimed to determine whether broadly speaking, the implementation of the EYLF (DEEWR, 2009) was reflected in improved outcomes in the general population of children. The AEDC is a population measure of how young children have developed by the time they start their first year of full-time school (Australian Early Development Census, 2023). Beginning in 2009, the AEDC has been undertaken every three years. If the implementation of the EYLF (DEEWR, 2009) had positive impacts on early developmental outcomes, improvements should be observed in AEDC outcome domains for children who attended ECEC over the period 2009 to 2021. We acknowledge that this approach is limited since we do not match individual children with ECEC centres. Nonetheless, population trends can be informative in evaluations of the considerable changes implemented in the ECEC sector over the past 15 years, and can inform the direction of future research and reform.

Background

Regardless of the complexity of the relations between ECEC quality and child outcomes, the prioritisation of access to quality education for all Australian children is underpinned by Target 4.2 of the United Nations Sustainable Development Goals (SDGs). Target 4.2 states that by 2030, all girls and boys should have access to quality early education and care that equips them for the transition to primary education (United Nations, 2015). However, in the absence of proposed indicators for quality in the SDGs, definitions of quality differ by country, perhaps due in part to differing stakeholder priorities (Cohrssen et al., 2023; Raikes et al., 2023). In Australia, national benchmarks are set in line with multiple quality areas (ACECQA, 2023a). The revised Early Years Learning Framework (EYLF; Australian Government Department of Education, 2022) and its previous iteration (Department of Education Employment and Workplace Relations (DEEWR), 2009) are intended to complement the NQS quality areas and thus support optimal outcomes for children. Constructs of quality are frequently categorised as structural or process, with structural quality (e.g., adult–child ratios, teaching qualifications, physical environment including materials and equipment) deemed to facilitate the enactment of process quality. Around the world, in addition to national regulatory assessment and rating processes, multiple observation-based measures of process quality focus on indicators of responsive adult–child interactions that support children’s social and emotional development, engagement in learning, concept acquisition and investigation and executive functions (Harms et al., 2015; Pianta et al., 2008; Siraj et al., 2015). Measures of structural quality are more easily observable and typically controlled by regulatory authorities such as accrediting organisations.

To date, the absence of fine-grained, population-level child attendance data has constrained efforts to assess the impact of ECEC attendance and centre quality. This is despite the critical importance of information on access to, and participation in, ECEC being essential for monitoring equity in access and participation (Raikes et al., 2023). Whilst the importance of including attendance data is recognised, this information tends to be binary (i.e. attended/did not attend). Large-scale Australian studies that have collected parent-reported ECEC attendance rely on data collected some ten years ago, preceding recent initiatives to increase enrolments and improve quality (Harrison et al., 2024). A recent analysis of 19 long day care (LDC) and preschool centres/schools in low socioeconomic communities showed wide variations in attendance and suggested that 49% of children in the study may have attended an ECE program for less than one year prior to starting school, and that attendance rates were higher for children in long day care than those in preschool (Harrison et al., 2024). This is important as it highlights the challenges for analysing outcome measures when using an attended/did not attend measure.

Indeed, while the impact of participating in ECEC does not necessarily improve child outcomes (van Huizen & Plantenga, 2018), the quality of ECEC has an impact: high quality has been positively associated with improved outcomes for some children (Melhuish & Gardner, 2021; Melhuish et al., 2015; Rankin et al., 2024; van Huizen & Plantenga, 2018). Nonetheless, a recent study refers to the ‘unsettled science’ on the long-term effects of early childhood education participation on sustained improvements to child outcomes and draws attention to the need for more rigorous research to investigate what works, and in particular, what works for children with disadvantaged backgrounds (Burchinal et al., 2024).

Aside from definitions of quality, access to ECEC remains a thorny issue. There are fewer ECEC services in low socioeconomic status areas of the country and those services tend to provide a lower average quality of care so children from more disadvantaged backgrounds are less likely to experience the quality of ECEC associated with developmental gains (Cloney et al., 2016). Even when these children are enrolled in ECEC, they are likely to attend fewer hours than their peers for reasons that include (but are not limited to) difficulty in affording ECEC fees, transportation challenges, poor fit between ECEC hours and caregiver working hours, and concerns that children will be exposed to illnesses (Beatson et al., 2022). Indeed, an examination of the differences in child development and SES of children within and between Australian states and territories using AEDC data from 2009 to 2018 found differences across jurisdictions: after adjusting for socio-demographic differences, developmental vulnerability declined over time in Western Australia and Queensland, remained stable in other jurisdictions but increased in the Australian Capital Territory (Collier et al., 2020). Australian research indicates that children who are at higher risk of developmental disadvantage are less likely to participate in ECEC (Taylor et al., 2022). Focusing specifically on Queensland, Areed and colleagues (2023) highlighted that ‘the complex spatial relationship between Indigenous status, preschool attendance and developmental vulnerability among children’ (p. 14) needs to be addressed in future work.

The Australian National Quality Framework (NQF)

In Australia, the NQF is designed to improve ECEC service quality by setting minimum national quality standards and assessing ECEC service against these standards. The NQF encompasses the National Law and National Regulations, the National Quality Standard (NQS), the assessment and quality rating process, approved learning frameworks (such as the EYLF), regulatory authorities across the states and territories that are responsible for the approval, monitoring and quality assessment of services, and the Australian Children’s Education and Care Quality Authority (ACECQA, 2023a). ACECQA works with regulatory authorities across the country and guides the implementation of the NQF.

The NQF has six stated objectives, two of which are:

  • Improve the educational and developmental outcomes for children attending education and care services;

  • Promote continuous improvement in the provision of quality education and care services (ACECQA, 2023a, p. 9).

The NQF has wide influence, covering most long day care (LDC), family day care (FDC), outside school hours care (OSHC) services as well as preschools/kindergartens, collectively known in Australia as early learning centres or early childhood education and care services (ECEC). The NQF introduced changes to structural quality such as improved educator-child ratios and minimum qualification requirements to increase opportunities for interactions that foster care and learning for individual children (process quality), and a quality rating system that is accessible to families indicating the quality of the service(s) in which their children are enrolled.

The NQS, as one element of the NQF, includes seven quality areas (QAs). These focus on services’ educational program and practice (QA1), children’s health and safety (QA2), the physical environment (QA3), staffing arrangements (QA4), relationships with children (QA5), partnerships with families and communities (QA6), and leadership and service management (QA7). Quality ratings include ‘Significant improvement required’, ‘Working towards NQS’, ‘Meeting NQS’, ‘Exceeding NQS’ and ‘Excellent’ (the latter rating can only be awarded by ACECQA). An NQS Assessment and Rating Instrument is used by authorised officers across the states and territories to facilitate the assessment and rating of services, and the first quality assessment visits were carried out in June 2012 (ACECQA, 2012). By 31 December 2013, 4,508 services had received a quality rating, representing 32% of all approved services at that date (ACECQA, 2014). Of the 4,508 services assessed, 12 received a ‘Significant improvement required’ rating and 1,811 received a ‘Working towards NQS’ rating.

In 2013, 59% of assessed services were Meeting or Exceeding NQS (ACECQA, 2014), whereas by 2021, when 93% of services had received a quality rating, 87% of services were rated as Meeting or Exceeding NQS (Table 1).Footnote 1 This indicates a marked increase in ECEC service quality as rated against the NQS. In addition, the proportion of services rated as Meeting or above NQS for QA1 increased from 67% in 2013 to 90% in Q4 2021. Notwithstanding this improvement, compared with other quality areas, QA1 (Educational program and practice) has remained the quality area with the lowest proportion of services Meeting or above NQS over that period (ACECQA, 2022).

Table 1 Proportion of services Meeting or Exceeding NQS, 2013–2021
Full size table

The Early Years Learning Framework for Australia (EYLF)

To enact ACECQA QA1, early childhood educators and teachers turn to the Early Years Learning Framework for Australia (AGDE, 2022; DEEWR, 2009) for guidance. The EYLF (AGDE, 2022; DEEWR, 2009) thus contributes to a national mechanism to ensure ECEC quality standards are met. Learning outcomes within the EYLF (AGDE, 2022; DEEWR, 2009) are child-centred and include:

Outcome 1: Children have a strong sense of identity.

Outcome 2: Children are connected with and contribute to their world.

Outcome 3: Children have a strong sense of wellbeing.

Outcome 4: Children are confident and involved learners.

Outcome 5: Children are effective communicators.

One of the stated aims of the EYLF is ‘to extend and enrich children’s learning from birth to five years and through the transition to school’ (DEEWR, 2009, p. 5) and a revised EYLF was released in early 2023 (Australian Government Department of Education, 2022). To the best of our knowledge, no assessment of the impact of the EYLF on child outcomes has been attempted, due to the absence of detailed data regarding children’s participation (as opposed to enrolment) in ECEC in Australia. In this paper, we focus on the association between the implementation of the first version of the EYLF (DEEWR, 2009) and multiple three-yearly rounds of child-level assessment data collected in the first year of formal education (Australian Early Development Census (AEDC), 2023).

The Australian Early Development Census (AEDC)

The AEDC is a population measure used to collect information from teachers every three years on approximately 100 items across five key domains of child development in the first year of full-time at school. It has been described as an ‘unparalleled resource for taking a national pulse of how Australia’s children are developing’ (Collier et al., 2020, p. 8). Box 1 shows the title and scope of each domain (AEDC Data Dictionary, 2022, p.4).

The first round of data collection occurred in 2009, with subsequent collection in 2012, 2015, 2018, and 2021. (Data collection is taking place again in 2024.) Data are aggregated and reported in group form. In response to concerns raised by Indigenous and non-Indigenous researchers, educators and other stakeholders regarding the need for AEDC processes to be culturally inclusive, an Indigenous adaptation of the AEDC was undertaken and the adaptation was validated using a multiple-stage, mixed methods approach (Silburn et al., 2009). Findings from this study informed the adaptation of the AEDC instrument and the process of its implementation in subsequent collection cycles.

Trends in AEDC data suggest that child development has improved in Australia since the first round of data was collected in 2009 (Harman-Smith et al., 2023). More specifically, from 2009 to 2021, the percentage of children assessed as developmentally on track shows a significant increase in all domains, and the percentage of children assessed as developmentally at risk has shown a significant decrease in all domains; nonetheless, the percentage of developmentally vulnerable children has shown a significant increase in physical health and wellbeing and social competence (Department of Education, 2022). The AEDC National Report 2021 also demonstrates that the percentage of children assessed as developmentally on track in five domains decreased slightly from 55.4% to 54.8% from 2018 to 2021 – perhaps reflecting the impact of the COVID-19 pandemic (Department of Education, 2022).

Harman-Smith and colleagues (2023) propose that shifts observed in 2012 may be attributable to reforms of the Universal Access National Partnership, NQS and EYLF. Indeed, a review of the National Partnership Agreement reported it to have been a major success in achieving an increase from 12% in 2008 to 96% in 2018 of children enrolling in a preschool program for the target 600 h (Royal Commission into Early Childhood Education and Care & Centre for Policy Development, 2023).

The Australian Productivity Commission uses two measures to define ECEC participation: first, ‘the proportion of children who are enrolled in Australian Government CCS approved child care service by age group’ and second, ‘Preschool program participation – the proportion of children who are enrolled in a preschool program in the state-specific YBFSFootnote 2’ (SCRGSP, 2024, p. 20). On average, children attended ECEC for 33.0 h per week in 2023 whereas average family day care attendance was 24.9 h per week (SCRGSP, 2024, p. 21). Nationally, 89.1% of children were enrolled in a preschool program in the state-specific YBFS in 2022. However, to be deemed enrolled, a child need only have attended the program for at least one hour during the reporting period, or be absent but expected to return, indicating the high level of variability of ECEC enrolment.

Multiple studies indicate that fewer children with language backgrounds other than English (LOTE), Aboriginal and Torres Strait Islander children, and children from lower socioeconomic status backgrounds attend ECEC than their peers (Brownell et al., 2016; Goldfeld et al., 2016; Harrison et al., 2024). However, whilst children from disadvantaged communities who attended some form of preschool before school entrance performed better than their peers who did not, children living in advantaged communities still achieved higher outcomes on the AEDC, and thus ECEC participation did not reduce the inequity associated with socioeconomic status (Goldfeld et al., 2016). A whole-of-population study in Australia suggests that differences in child outcomes may be explained to some extent by the fact that ECEC services in regional and remote areas are more likely to be rated as standard quality (Meeting NQS) or low quality (Working towards NQS or Significant Improvement Required) rather than high quality (Exceeding NQS or Excellent; Tang et al. 2024). Additionally, prior research has shown that factors including parental education, socioeconomic background, and geographical location predict AEDC domain outcomes (Guthridge et al., 2016; Taylor et al., 2020).

Across the country, efforts continue to be made by governments to increase access and by far the majority of children aged three to six years attend centre-based care (ACCC, 2023). Calls for improved measurement of children’s ECEC attendance persist (Cohrssen et al., 2023; Gilley et al., 2015; Nous Group, 2020). Perhaps due in part to the absence of detailed information regarding child participation in ECEC programs, 14 years after the introduction of the EYLF (DEEWR, 2009), its impact on child outcomes has not been extensively assessed. However, whether positive impacts of the EYLF (DEEWR, 2009) can be observed in the population and in developmental domains other than literacy, is unknown. In theory, if the increases in NQS QA1 (Educational Program and Practice) impact child outcomes, then these gains should be observed – at least to some extent – at a population level in the AEDC data collected for the cohorts that follow the 2009 cohort which coincided with the rollout of the EYLF (DEEWR, 2009), i.e. 2012, 2015, 2018 and 2021 waves.

Method

Full university ethics approval was obtained before commencing data analysis (HE23-136). We examined whether improvements in child outcomes on the five AEDC domains (physical health and wellbeing, social competence, emotional maturity, language and cognitive skills, and communication and general knowledge) can be observed for children reported to have participated in ECEC over the five waves of AEDC data collection (Janus et al., 2011).

To achieve this, we mapped mean scores on each of the five domains from the initial data collection (2009) to the latest available data (2021). Since this study focuses on formal ECEC attendance (as opposed to in-home or family day care) child participation was indicated by a yes/no response to Attended a day care centre, and/or Attended a pre-school program. Since the EYLF was rolled out in 2009 (the first round of AEDC data we included), and we were interested in high-level impacts of the EYLF generally, children were categorised as ‘attending’ if they attended either a day care centre or a preschool program, or both.

We compared ECEC participants with non-participants at each wave since children participating in ECEC should be more likely to demonstrate increased learning outcomes when compared with non-participating children.

The primary research question in this study therefore is:

1. Do trends in means on each of the five AEDC subdomains differ for children who attended vs did not attend ECEC programs in the year before school entry?

We additionally address the following sub-questions:

2. Has the proportion of children attending vs not attending ECEC changed over the five rounds of data collection?

3. Do attendance patterns differ by geographical location over time?

4. Do trends in means on each of the five AEDC subdomains change over the five rounds of data collection?

5. Do year of attendance, location, sociodemographic factors and ECEC attendance together predict AEDC domain outcomes?

Measures

ECEC Attendance. Teachers completed the ECEC attendance variables as part of the questionnaire for each child (Department of Education, Skills & Employment [DESE], 2022). ECEC attendance was computed by combining responses to the questions, To the best of your knowledge, has this child been in the following forms of non-parental care on a regular basis in the year before entering school (2009/2012) or To the best of your knowledge, did the child attend a preschool/kindergarten program in the year before entering full time school? (post-2012 version). To ensure the data were as comparable as possible over the five data collection waves we used the derived items computed by the AEDC from these questions, indicating whether a child attended either preschool or daycare. A three-level categorical variable was calculated to comprise ‘yes’, ‘no’ and ‘don’t know’ responses. A child attending any form of day care or preschool, either part-time or full-time was considered as attending (i.e. coded as ‘yes’). In this study, we consider attendance at formal ECEC as a proxy measure for exposure to the EYLF.

Geographical Remoteness was a five-level categorical variable indicating the location of the school the child attended. The remoteness categories (Major cities, Inner regional, Outer regional, Remote, Very remote) were based on the Australian Statistical Geographical Standard Remoteness Areas and are comparable over each wave of data collected (DESE, 2022).

Socio-Economic Indexes for Areas (SEIFA) represents socioeconomic advantage and disadvantage of geographical areas within Australia. The measure was developed by the Australian Bureau of Statistics using census data. Higher SEIFA values represent more socioeconomically advantaged areas (DESE, 2022). Both SEIFA and Geographical Remoteness variables were collected at the school level (i.e. not completed by questionnaire respondents).

Age is an ordered categorical variable based on a student’s exact age at the time of the AEDC assessment; Gender is a binary variable indicating male or female. Both variables were collected as part of standard school enrolment processes, with the Age category computed using students’ birthdates and the date of questionnaire completion (DESE, 2022).

AEDC domains. The AEDC assessment is an Australian version of the Early Development Instrument (AvEDI), initially trialled and refined in Canada (Janus & Offord, 2007; Janus et al., 2011). The Australian version was developed so that it was specifically relevant to the Australian context (Andrich & Styles, 2004). The instrument has demonstrated concurrent and predictive validity, and internal and inter-rater reliability (Gregory et al., 2021). The 96 items in the AvEDI instrument are completed by classroom teachers for each child during the first year of full-time school. Items are either 3- or 5-point Likert-style questions and data is entered directly by teachers into an online form. Items are then combined to form five domain scores. Titles of the scope of each domain are in Box 1. Continuous domain score variables are then categorised and proportions of the population are labelled as developmentally Vulnerable, Developmentally At-Risk, or Developmentally On Track. Results of each AEDC data collection round are reported as proportions of children in each of these categories and combinations thereof (e.g. vulnerable in all five domains). Given the rigorous psychometric development of the AvEDI domain indicators using a Rasch modeling procedure (Andrich & Styles, 2004), in this study, we use domain scores as continuous variables rather than applying or using the developmental categorisations. Higher domain scores indicate better overall ratings on each outcome.

Five domains of early childhood development measured by the Australian Early Development Census (AEDC)

The AEDC measures five areas or ‘domains’ of early childhood development from0020information collected through a teacher-completed instrument:

• physical health and wellbeing – measures children’s physical readiness for the school day, physical independence and gross and fine motor skills

• social competence – measures children’s overall social competence, responsibility and respect, approaches to learning and readiness to explore new things

• emotional maturity – measures children’s pro-social and helping behaviour, anxious and fearful behaviour, aggressive behaviour and hyperactivity and inattention

• language and cognitive skills (school-based) – measures children’s basic literacy, interest in literacy, numeracy and memory, advanced literacy and basic numeracy

• communication skills and general knowledge – measures children’s communication skills and general knowledge

AAA

  1. Note From Australian Early Development Census (AEDC) Data Dictionary (DESE, 2022).

Data analysis plan

As noted above, the AEDC reporting on ECEC attendance enabled us to split the sample into three distinct categories for each year of data collection: (1) those children who attended preschool or day care (or both) in the year before commencing school; (2) those children who did not attend preschool or day care; and 3) those for whom ECEC attendance was unknown. We retain category (3) (unknown) for all analyses since dropping these children would introduce avoidable bias; that is, it is not known if these missing responses can be classified as ‘missing at random’ (MAR) or ‘missing completely at random’ (MCAR), so retaining them in the sample is recommended, particularly for regression models (Enders, 2013).

To examine ECEC attendance over time, we first calculated absolute numbers (n) and proportions of children attending and not attending ECEC within each AEDC reporting year. We also calculated n and attendance proportions by geographical location (Major cities, Inner regional, Outer regional, Remote, Very remote). We present results in both tables and figures for ease of interpretation.

To observe changes in AEDC domain average scores we first present time series figures for each domain separately. Next, to examine average differences between ECEC attendees and non-attendees, we present time-series plots of domain average scores by ECEC attendance group, retaining the ‘not known’ group in all plots. Finally, we include a table showing the mean score for each domain at each wave of data collection, and the difference scores between the ECEC attendees and non-attendees.

To evaluate the extent to which attendance year, socio-demographic factors, and ECEC attendance predicted AEDC domain outcomes, we fitted a series of hierarchical regression models. Using each domain score as an outcome variable, we included predictors in three sets. The first set of predictors included binary indicators of the year of data collection. We used 2009 as the comparison year since it was the first year AEDC data were collected. In this model, the coefficient for each year represents the average difference between 2009 and subsequent years. Including year as a binary variable also allows us to control for any cluster effects (i.e. these are fixed effects regression models, McNeish et al., 2017), so that coefficients of predictors in subsequent steps can be interpreted as effects that have previously controlled for changes over time. The second set of predictors included socio-demographic factors known to relate to early childhood developmental outcomes. These include socioeconomic status (indicated by the SEIFA measure), age, gender, and residential location. The final model includes ECEC attendance as a three-level categorical predictor, with ‘did not attend’ as the base category.

Given that the AEDC is a survey of the population of children, we are mindful that small effects are likely to be statistically significant while potentially not being practically meaningful (Cumming, 2014). We therefore report both unstandardized (B) and standardised (β) regression coefficients, and confidence intervals for standardised coefficients. Standardised regression coefficients are also interpreted as effect sizes in standard deviation units (SDs; essentially the same interpretation as a Cohen’s D effect size statistic). The R2 statistic for each step of the hierarchical regression models is also reported and indicates the proportion of variance in the AEDC outcomes explained by the sets of predictors in each step of each model.

Results

Figure 1 (top left panel) shows the proportions of Australian children attending any form of ECEC for the five waves of AEDC data collection. ECEC Attendance is further broken down by geographical location in Fig. 1 (remaining five panels). Table 2 reports the specific percentages for each wave, location and the total size of each cohort. The proportion of Australian children reported to have attended ECEC in the year preceding the start of full-time school increased from 81.8% in 2009 to ~ 86% in both 2018 and 2021. The proportion of children not attending declined across this same period from 10.1% in 2009 to 5.6% in both 2018 and 2021. There remains a proportion of children for whom this information is not available in each year (7.9–10.7%). As expected, ECEC attendance was highest in Major Cities and Inner Regional areas of Australia. Participation rates were lower in Outer Regional, Remote and Very Remote locations; however, there was an increase in the proportion of children attending any form of ECEC over time in all locations. For example, 66% of children in Very Remote locations were reported to have attended ECEC in 2009, with 20% not attending. This increased to 81% of Very Remote children attending ECEC in 2021, with 8.5% not attending.

Fig. 1
figure 1

Early childhood education and care (ECEC) attendance overall (top left panel) and by geographical remoteness from 2009 to 2021 bars indicate the proportion of children in the population attending ECEC (‘Yes’), not attending (‘No’) and those for whom attendance was unknown (‘Don’t know’)

Full size image
Table 2 Sample Sizes (N) and Early Childhood Education and Care (ECEC) Attendance Percentages by Remoteness Category and Total Population 2009—2021
Full size table

Time series mean score plots for each of the five AEDC domains are shown in Fig. 2. Each domain score has a range of 1–10, though we scale the y-axes for each domain from 5–10 in these and subsequent plots. The physical health and wellbeing domain had persistently the highest average of ± 9 on the 1–10 scale, with minimal change over the six years of data collection. Social competence and emotional maturity also showed minimal change in the population mean over time, with average scores ranging between 8 and 8.5 on the 10-point scale. Language and cognitive skills (school-based) showed an improving trend from 2009 to 2018, with a slight decline from 2018 to 2021, while communication skills and general knowledge showed slight improvements.

Fig. 2
figure 2

Mean scores for five domains of the Australian Early Development Census (AEDC) in the Australian Population 2009–2021

Full size image

As part of the data checking procedure, and to provide additional information about the domain indicators to be used as dependent variables in regression models, we generated histograms for each of the five AEDC domains. Supplementary Fig. 1 shows these distributions. All five domain outcome variables were notably negatively skewed, indicating that the majority of children were positively rated on the AEDC instruments, with far fewer children rated poorly. While negatively skewed outcome variables can bias standard errors in regression models, and hence p-values (Altman & Bland, 1995), since we have access to data from the population of children in this study (i.e. it is not necessary to generalise from a smaller sample to the population in these analyses), the impact of this bias is mitigated. Nonetheless, we rely more on interpretations of effect sizes rather than statistical significance in the reporting of regressions below.

To examine whether differences in average scores on each of the five domain outcomes differed by whether children had attended ECEC or not, Figs. 3, 4, 5, 6, 7 were generated. These figures show the time series of average scores for the three groups identified above: those who were recorded as attending ECEC, those who did not, and those for whom this information was unknown. In all five domains, the average scores for the group who attended ECEC were consistently higher than those for the group who did not attend ECEC. Table 3 shows mean scores by domain and data collection year, and the difference score between those attending and not attending ECEC. For example, in the emotional maturity domain, average differences between ECEC attendance groups were quite small (from 0.27 to 0.38 points on the 1–10 scale). Physical health and wellbeing, and social competence showed slightly larger average differences: 0.44 to 0.65 points for the former, and 0.47 to 0.67 points difference for the latter. The largest differences between the group attending ECEC and those not attending were in the language and cognitive skills (1.00 to 1.19 points), and communication and general knowledge (1.09 to 1.51 points). Interestingly, increases in difference scores over the five waves appear to be due to the ECEC non-attending group averages declining rather than the ECEC attending group averages increasing. The possible exception is the Language and Cognitive Skills domain, though this data is characterised by a curved, trend which is difficult to interpret. The group for which ECEC attendance information was not known, consistently had average scores between the ‘attended’ and ‘did not attend’ groups in all five domains, suggesting this group was generally made up of a mix of children who did or did not attend (i.e., this was not a specific population of children different from others whose information was captured).

Fig. 3
figure 3

Physical health and wellbeing domain mean scores by ECEC attendance group 2009–2021

Full size image
Fig. 4
figure 4

Social competence domain mean scores by ECEC attendance group 2009–2021

Full size image
Fig. 5
figure 5

Emotional maturity domain mean scores by ECEC attendance group 2009–2021

Full size image
Fig. 6
figure 6

Language and cognitive skills domain mean scores by ECEC attendance group 2009–2021

Full size image
Fig. 7
figure 7

Communication skills and general knowledge domain mean scores by ECEC Attendance Group 2009–2021

Full size image
Table 3 Mean Scores for each AEDC domain by ECEC attendance group and difference scores 2009–2021
Full size table

While Figs. 3, 4, 5, 6, 7 are suggestive, interpreting average group differences in a naive way cannot add (and may obscure) additional information about other factors that may be important in influencing early childhood developmental outcomes, over and above attendance at ECEC. We therefore next report a series of five hierarchical regressions that seek to examine the extent to which several factors predict AEDC domain outcomes, in addition to any effects of attending ECEC.

Table 4 shows results for the physical health and wellbeing outcome. The first step of this model, including only year of data collection as a series of binary predictors (with 2009 as the comparison year), showed that average changes in the population over the five rounds of the AEDC data collection were negligible. The R2 for this model was effectively zero, indicating that none of the variance in the physical health and wellbeing outcome was explained by year of assessment. The second step included demographic and background factors as predictors, while retaining the year-indicator variables. Children from higher SEIFA areas had higher scores on average on the outcome variable; older children also had higher ratings compared with younger children, and girls were rated higher than boys on the physical health and wellbeing outcome. For geographical remoteness, Major Cities was included as the reference category, with the remaining four geographical locations compared with the referent. Compared with Major Cities, Inner Regional, Outer Regional and Remote located children had slightly lower average ratings on the outcome. The Very Remote group had on average, the lowest scores on the Physical Health and Wellbeing outcome compared with the Major Cities group (0.33SD). Including this set of predictors explained 4.5% of the variance in the physical health and wellbeing variable.

Table 4 Hierarchical regression model results for physical health and wellbeing domain
Full size table

The final step retained both sets of predictors in the previous two steps, and included a categorical predictor of ECEC attendance. Given the proportions of each sample where teachers indicated they did not know whether a child had attended ECEC, we have retained this group in the analyses to avoid potential biases due to missing data. Both the ‘unknown’ group and the ‘attended’ group are compared with the ‘did not attend group’ (the reference category). On average, children who had attended ECEC had 0.32SD higher ratings on the physical health and wellbeing outcome compared with children who did not attend. However, the addition of these predictors explained < 1% of additional variance in the outcome.

Tables 5, 6, 7, 8 show the results for the same series of hierarchical regressions for the four remaining AEDC outcomes. The coefficients reported in each can be understood similarly to those for the physical health and wellbeing domain. For the sake of brevity, we highlight here notable similarities and differences with the results for the first domain described above. First, and as might be expected from Figs. 3, 4, 5, 6, 7, only the language and cognitive skills domain showed meaningful average differences between 2009 and the subsequent years. For example, the 2012 group had 0.15SD higher average ratings on language and Cognitive skills compared with the 2009 group. The average was ~ 0.24SD higher in the 2015 and 2018 groups, and back to 0.15SD higher in the 2021 group – tracing the curve of improvement to 2018 and a slight decline between 2018 and 2021 that can be seen in Fig. 2. Notwithstanding these trends, only 1% of the variance was explained by the inclusion of the binary year indicators for this outcome. The remaining three domains showed minimal average changes by year of data collection. Nonetheless, we retain the year indicator in the remaining models since these data are clustered into five different cohorts.

Table 5 Hierarchical regression model results for social competence domain
Full size table
Table 6 Hierarchical regression model results for emotional maturity domain
Full size table
Table 7 Hierarchical regression model results for language and cognitive skills domain
Full size table
Table 8 Hierarchical regression model results for communication skills and general knowledge domain
Full size table

In the second step of the models, SEIFA was a consistent predictor of each outcome (β = 0.11–0.17), as was age (β = 0.05–0.15). Comparing across domains, age was the strongest predictor in the language and cognitive skills domain. Gender was also a reliable predictor of each of the domain outcomes, with girls having higher ratings on average than boys (β = 0.12–0.26). Of particular note is the finding that girls were on average approximately 0.20SD above boys on social competence and 0.25SD above boys on emotional maturity, even after controlling for age, location, and SEIFA. In general, there were minimal differences in all domain outcomes between children located in Major Cities, and Inner or Outer Regional and Remote areas (all βs < 0.10). An exception here was the finding that children in Remote areas had 0.26SD lower average ratings on the language and cognitive skills domain compared with children in Major Cities. Children in Very Remote areas had consistently poorer ratings in all domains compared with children in Major Cities, ranging from 0.30SD (communication skills and general knowledge) to 0.33SD (social competence) and 0.34SD (emotional maturity). The largest difference was observed for the language and cognitive skills domain with Very Remote children on average 0.72SD below their Major Cities peers even after controlling for SEIFA, age, gender and year of data collection.

The final step retained all predictors and included the ECEC attendance binary variables. Again, similar to the results reported for the physical health and wellbeing domain, all remaining domains showed that children who attended any form of ECEC had consistently higher ratings on the remaining four outcomes. The size of the average differences varied by domain, with the smallest differences in the emotional maturity domain (β = 0.13), followed by social competence (β = 0.24). The largest differences between children attending and not attending ECEC were observed in the communication skills and general knowledge domain (β = 0.45) and the language and cognitive skills domain (β = 0.48). In other words, children attending ECEC had on average almost half a standard deviation higher ratings on these latter two domains, even after controlling for year of data collection, SEIFA, age, gender and geographical location. Notwithstanding these results, the R2 was small in all models, with only 5–10% of the outcome variance explained by all predictors combined.

Discussion

This study aimed to examine whether the implementation of the EYLF (DEEWR, 2009) co-occurred with improvements in developmental outcomes at the population level for children who attended ECEC, given increased attendance and increased quality as assessed against the NQS. We mapped early childhood developmental outcomes measured by the Australian version of the Early Development Instrument over the years following the roll-out of the EYLF (DEEWR, 2009) – the same period over which improvements in ECEC quality have been documented (i.e., 2009 to 2021). Using population datasets in the way we have in this study provides preliminary evidence about the impact of quality improvements on child outcomes at a population level. What these data lack in precision, is somewhat mitigated by coverage, given that entire cohorts of children are assessed via the AEDC every three years, and all ECEC centres are assessed against the NQS.

As noted in the introduction, the percentage of ECEC centres meeting or exceeding the NQS increased from 59% in 2013 to 87% in 2021, demonstrating the large-scale improvements in ECEC quality over this time. Relatedly, since 2017, 93% to 94% of all ECEC services have been captured in the NQS rating process. This reflects an enormous effort and investment, and has resulted in improved ECEC quality according to the standards set by the NQF (ACECQA, 2023b). Concurrently, initial results of this current study show that increasing proportions of Australian children have attended ECEC in the year preceding the start of full-time school: 82% in 2009 increasing to 86% in 2021. Increases in attendance rates were also observed in all geographical locations, although children in Very Remote locations had the lowest overall attendance by 2021 (81%), and children in Inner Regional and Major Cities had the highest (88 and 86% respectively). Theoretically, increasing percentages of children attending higher quality ECEC should be reflected in gains in developmental outcomes in the population of children who attend care or preschool in the year before beginning formal education (von Suchodoletz et al., 2023).

To observe whether any such relation was observable in population data, we generated average scores on the five AEDC domains over the five years that data were collected (every three years since 2009). While AEDC outcomes are most often reported as categories of children ‘developmentally on track’, ‘developmentally at risk’, or ‘developmentally vulnerable’ (Department of Education, 2022), the AEDC domain measures were designed to provide continuous scores in a distribution (Andrich & Styles, 2004). Given the same survey is completed at each assessment for different population cohorts, average changes over time can be interpreted. In each of the five AEDC domains, a persistent gap was evident in average scores, with children attending any form of ECEC rated more highly on average over every round of data collection. This gap differed in magnitude, with the greatest difference in language and cognitive skills, and communication skills and general knowledge, with the smallest group differences in social competence and emotional maturity. Interestingly, these group average differences neither widened nor narrowed substantially from 2009 to 2021, with the gap remaining remarkably consistent in each of the five domains. Small increasing difference-scores between ECEC attending and non-attending groups were largely due to the averages declining for the non-attending groups rather than increasing for the attending groups. That is, notwithstanding the introduction of the NQF and documented improvements in ECEC quality as indicated by improvements when assessed against the NQS, these improvements in ECEC offerings are not observable at a population level in early childhood developmental outcomes for those children attending ECEC.

Regression models for each of the five AEDC outcomes support and extend the conclusions drawn from patterns in average scores and group differences over time. In particular, ECEC attendance remained a predictor of each of the five developmental outcomes even after multiple demographic and locational variables were included in the models. Nonetheless, attendance at ECEC explained minimal variance in each model: from 0.2 to 2% of the variance in each outcome was explained by the categorical ECEC attendance variable. By contrast, including SEIFA, gender, age and geographical location together explained between 4.5 and 8% of the variance in each of the five outcomes. These results reflect prior research examining predictors of early developmental outcomes (e.g. Guthridge et al., 2016; Taylor et al., 2020). Additionally, small or no impact of ECEC attendance on childhood development are not unknown in the literature (Burchinal et al., 2024; van Huizen & Plantenga, 2018).

What can we make of these results? First, it is possible that the features of quality ECEC that are guided by the NQS and the EYLF (DEEWR, 2009) – and which are therefore the focus of the efforts of educators – are those that may not alter the developmental trajectories assessed by the AEDC in a meaningful way in the population. That is, if early learning and development and, by extension, later childhood outcomes are to be improved, interventions should be aimed at developmental domains – or learning areas – that are malleable, fundamental to later progress, and would not have developed in the absence of improving ECEC quality (Bailey et al., 2020; Burchinal et al., 2024). In other words, it is clear that participating in ECEC does something to improve early childhood outcomes since attendees are ahead relative to non-attendees, but what aspect of ECEC has this effect is less clear since differences between groups do not appear to be caused by developmental outcome improvements for children attending ECEC, even while ECEC quality has improved. It is possible that the value of ECEC may not be related to the quality domains that are currently measured by the NQS (e.g., Burchinal et al., 2021).

Second, these findings highlight the need for more fine-grained child participation data, as well as longitudinal data relating to teacher qualifications and teaching quality, so that rigorous research may identify and evaluate aspects of quality that have a positive and sustained impact on child learning outcomes. Third, the much-anticipated Preschool Outcomes Measure currently in development is intended to assist teachers and educators to understand children’s learning and development against two domains: executive function and oral language and literacy (Australian Government Department of Education, n.d.). This measure will provide supplementary information to the AEDC and importantly will be measured in the preschool years. However, it will not contribute to an evaluation of the impact of the EYLF (ADGE, 2022; DEEWR, 2009) on improved outcomes for children.

Limitations

We acknowledge that our analyses are limited and make no attempt to claim representativeness or depth of measurement. Notwithstanding the rigorous development of the AEDC, the univariate distributions for each of the domains reported in Supplementary Fig. S1 suggest ceiling effects in the measures. It is possible that the AEDC instrument is not sensitive enough to capture the full range of individual differences in each outcome, and this information should be considered in the use of the scale in future research.

Secondly, the absence of detail regarding children’s participation rates in ECEC within the AEDC is problematic. Attendance is presented as a binary ‘attended’ or ‘did not attend’, masking the wide variation in attendance. Measures of enrolment in ECEC may not reflect actual participation in ECEC when enrolled, and the absence of this information may skew findings at the population level and population-wide child outcomes data is essential for accurate monitoring and evaluation of policy impact (Goldfeld et al., 2016; Raikes et al., 2023).

A third limitation of this study is that no data are available regarding the quality of the ECEC services attended by children. However, multiple studies have indicated that high-quality ECEC is required to achieve gains in child outcomes, particularly for children with disadvantaged backgrounds (e.g., Melhuish & Gardner, 2021; Melhuish et al., 2015; Sylva et al., 2006; Tayler, 2016; van Huizen & Plantenga, 2018).

A fourth limitation relates to teaching qualifications. Some evidence of the positive impacts of early childhood teaching qualifications on primary school literacy outcomes in Australia has been reported (Warren & Haisken-DeNew, 2013) and a systematic review of 48 international studies reported positive correlations between qualifications and overall ECEC quality (Manning et al., 2017). Investigating the impact of teaching qualifications on quality and child outcomes could be another specific focus for future research examining the impact of ECEC attendance on early developmental outcomes.

Conclusions

Findings from three important studies have recently been published in Australia. One has focused on AEDC data and reports that children starting school are increasingly culturally diverse, that children from culturally diverse backgrounds are less likely to attend ECE, and these children are more likely to be developmentally vulnerable at school entry (Lam et al., 2024). A second study explored whether the quality of EC services attended by children predicted the children’s developmental outcomes in the first year of school, using the NQS as the measure of quality and incorporating E4Kids study data to provide detailed measures of quality and AEDC data as a measure of child outcomes (Rankin et al., 2024). Importantly, this study found that a rating of Exceeding the NQS is the threshold for reducing developmental vulnerability, and emphasised the crucial role played by teachers and educators in promoting children’s learning and development. A third study addressed a gap in extant research by focusing on enrolment and attendance records to examine child attendance patterns across the preschool year in areas of socio-economic disadvantage. This study reported wide variability in children’s hours of attendance (Harrison et al., 2024). Attendance patterns were more stable for children who attended preschools than those attending long day care (although attendance rates were lower in preschools than in long day care), and 49% of children in their sample may have attended ECE for less than one year before entering school – the time at which AEDC data are collected.

Our analysis of AEDC data adds to this discussion by highlighting the importance of continuing to improve sources of data that help evaluate the impact of improvements made to ECEC services on children’s outcomes. This priority has been confirmed by the Productivity Commission Inquiry Report (2024) which also proposes multiple strategies to reduce workforce challenges and address barriers to children accessing ECEC. We acknowledge the limitations of our approach. Population trends are worthy of examination in light of the considerable gains made by the ECEC sector over the past 15 years, and our findings can inform the direction of future research. It is important to acknowledge the progress towards universal access and the increased quality of early childhood education and care services. However, much more information regarding dosage in the form of children’s actual ECEC attendance than the current yes, no, don’t know responses, and the duration of attendance, would enable more nuanced findings and recommendations.

This conclusion is not new: the need for more detailed data has been recommended in multiple documents over many years (Nous Group, 2020) and work has commenced toward the development of an Australian Preschool Outcomes Measure that will support evidence-based assessment for learning and has the potential to generate more nuanced national child outcomes data. Moving forward, such findings may demonstrate the positive impact of enacting the revised Early Years Learning Framework (AGDE, 2022) at a population level over time, particularly in light of the addition of new practice principles (ACECQA, 2023b) that respond to societal priorities. Alternatively, findings may indicate the need for enhanced professional learning for early childhood teachers in intentionally facilitating children’s learning and development within and across multiple domains that are assessed when children enter full-time school.