An area-based measure of risk of social exclusion for Australian school-age children.
Miranti, Riyana ; Daly, Anne ; Tanton, Robert 等
1. INTRODUCTION
There has been much discussion in recent years about the two speed
economy generated in Australia by the mining boom and its effects on
economic outcomes for individuals living in different areas of the
country. While some States and specific areas within them have
prospered, notably in Queensland and Western Australia, other areas such
as Tasmania and New South Wales have experienced much lower rates of
growth. The aim of this paper is to present results for an index of risk
of social exclusion for school-aged children calculated at the small
area level using data from the 2006 Census, at the height of the mining
boom. In addition this paper will consider how children in different
states were faring at a time when there were significant differences in
the performance of state economies. Gross State Product (GSP) in
Queensland and Western Australia grew at over twice the rate of GSP in
the more populous states of New South Wales and Victoria between 2005
and 2006. In 2006, Western Australia had one of the highest real Gross
State incomes per capita while Queensland fell in the middle of the
distribution. The question of interest here is whether these differences
were reflected in different outcomes for children.
Social exclusion is a broader concept of disadvantage than income
poverty and recognises that non-participation can develop in various
ways such as discrimination, cultural identification, geographical
location or transport accessibility (Burchardt et al., 2002). Thus,
while the concept of poverty refers to a level of material resources
(i.e. whether the individuals live below or above the poverty line),
Saunders et al. (2008: 178) have argued that features of social
exclusion are not only associated with individual characteristics but
also with community, social and spatial characteristics. The phenomenon
has featured in the social discourse of many countries, with the
realisation from researchers, practitioners and policy makers that
disadvantage often spans many dimensions and phases of an
individual's life, including childhood. Social exclusion is one of
the measures that has been developed to describe multidimensional
poverty and disadvantage (for a discussion of related measures see
Alkire and Santos, 2013).
Social exclusion has implications for an individual's current
standard of living and also impacts on future quality of life. Levitas
et al. (2007) consider three further aspects of social exclusion. Wide
social exclusion arises where a large number of people are disadvantaged
on one indicator; deep social exclusion occurs where individuals face
multiple disadvantages on a range of indicators; and concentrated social
exclusion occurs where there is a geographical concentration of
disadvantage. Some individuals may experience all these aspects of
social exclusion.
The concept of social exclusion has been developed in the context
of children (see Ben-Arieh, 2008 and for discussions in the British and
European contexts, see for example Bradshaw et al., 2006; Levitas et
al., 2007 and Oroyemi et al., 2009). While there is limited reference to
the term 'social exclusion' in the American literature, many
of the same ideas have been used to develop indicators of child
wellbeing and disadvantage (see for example Land and Crowell, 2010).
Social exclusion for children has also been developed and constructed by
several Australian researchers who examined this concept in a spatial
setting (see for example Daly et al., 2008; McNamara et al., 2009 and
Tanton et al., 2010). The other side of social exclusion is social
inclusion which emphasises the positive aspects of children's lives
and many of the indicators identified mirror the exclusion indicators
and fall within the same domains. It has been argued that positive
outcomes for children involve more than just the absence of negative
indicators.
The aim of this paper is to bring together the concept of social
exclusion in the context of children and Australian regions. Expanding
previous work by Daly et al. (2008); McNamara et al. (2009), Tanton et
al. (2010) and Abello et al. (2012), this paper describes further the
refinement and analysis of the spatial index of social exclusion for
children in Australia (CSE Index). This work also adds to previous
studies which discuss other measures of child wellbeing in Australia for
example child poverty (Khan et al., 2000) and children's own
perspective on their wellbeing (Ramsden, 2013).
We have restricted the analysis here to focus on school-aged
children because while there are common issues relevant to all children,
there are particular variables relevant to this age group. For example,
educational outcomes and the role of connectedness take on particular
significance for school-aged children compared to pre-school children.
Limiting the study to school-aged children also enables the utilisation
of particular data relevant to them. In addition, school-aged children
are an important focus of education and health policy and there are a
range of policy instruments directed toward this group. Further work has
developed an index for pre-school aged children.
In this paper, the small area rather than the individual is used as
the unit of analysis. There are several reasons why the location of
disadvantage is important. From a practical and policy point of view,
there is often a geographical dimension to delivering policies, for
example in the areas of health and education. There may be significant
advantages in providing local communities with the information they
require to actively engage in promoting the welfare of the population in
their areas (Coulton et al., 2009 for the case of children). Miles et
al. (2008) specify the importance of evidence based information
regarding community and regional wellbeing as a superior mechanism to
inform and benefit not only the decision makers, but also the
communities. Erbstein et al. (2013) argue that there is evidence in the
literature of neighbourhood effects on the wellbeing of young people
with concentrations of high economic wellbeing being associated with
positive outcomes in young people (see Brooks-Gunn et al., 1993).
The literature has discussed a growing number of small area
disadvantage or wellbeing indicators for particular groups of the
population which have improved the ability of policy makers and
communities to target areas of disadvantage. Noble et al. (2010: 294)
proposed that having area-based measures will have three main objectives
from a policy perspective.
"Allocating resources and informing detailed service planning
by national, provincial and local governments (thereby increasing
transparency and accountability), and reducing the use of anecdotal
evidence that is not evidence-based. Policy related and academic
research (e.g. a sampling frame for in-depth studies or pilot studies; a
tool for contextualising other empirical research). Targeting resources
provided by donor agencies, companies, voluntary bodies and
charities".
As the information is typically provided at the level of the domain
as well as at an aggregate level in a comprehensive index, it is
possible to highlight the domains in which each area is most vulnerable
(see for example Barnes et al., 2009, Bradshaw et al., 2009, Land and
Crowell, 2008 and the Annie E Casey Foundation, 2010).
The paper is organised as follows. The next section discusses the
advantages of an area-based social exclusion concept for children. The
third section discusses the construction of the Child Social Exclusion
Index (CSE) for Australia. The fourth section examines and analyses the
results, followed by a concluding section.
2. THE ADVANTAGES OF AREA BASED SOCIAL EXCLUSION CONCEPT ON
CHILDREN
There are at least three advantages in adopting the concept of
social exclusion for children. Firstly, the multidimensional measure has
the advantage of recognising a broader range of indicators of
disadvantage beyond income and including the rights of the child, as
articulated in the UN Convention on the Rights of the Child. Further, it
addresses the concept of multidimensional disadvantage as developed in
Sen's capability framework (Sen, 1999). Sen (1999) indicated the
importance of a person's actual capability such as the ability to
be healthy and the ability to participate in community and society which
relies on a combination of factors, including physical and mental
characteristics and opportunities and influences to enable them to
actively participate (Cassells et al., 2011).
A second advantage of a broader measure of disadvantage is that it
focuses more on the process or potential risk factors rather than an
outcome such as income poverty. The CSE index includes many variables
which are indeed process variables associated with poverty. This means
the characteristics measured in the index increase the risk of being
excluded which may be associated with income poverty and other long-term
disadvantages. For example, the CSE index includes variables such as the
proportion of children with no parent in paid work (either unemployed or
not in the labour force), the proportion of children who live in a
single parent family, and the proportion of children who live in a
household where no members in the family have completed year 12. These
variables are linked to intergenerational transfer variables, so
children who live in families where their parents or other adults in
their family are socially excluded also experience a higher risk of
being excluded. This may then impact on the development of the children.
Finally, not only the individual is important, the
community/neighbourhood is also critical in determining outcomes. Lupton
and Power (2002) argue that social exclusion is affected by the nature
of the neighbourhood through three factors. A neighbourhood's
characteristics including location, transport availability, housing and
economic structure tend to be stable and difficult to change. There is
natural residential sorting which pools the population from the most
disadvantaged groups in the most disadvantaged neighbourhoods. Finally,
once the concentration of disadvantage is established, these
disadvantaged areas may acquire more harmful characteristics. For
example the areas can be characterised by a high level of crime, poor
health outcomes and other chronic measures of disadvantage. People
living in those neighbourhoods may face further difficulties such as
access to transport and find themselves stigmatised by their location of
residence when attempting to engage with the wider community for
employment and social activities. There has been much research on these
neighbourhood effects (see Brooks-Gunn et al. (1993) and Goldfeld et al.
(2014) for a recent review of the literature).
3. CONSTRUCTION OF THE CHILD SOCIAL EXCLUSION (CSE) INDEX 5-15 IN
AUSTRALIA
Data and Spatial Unit of Analysis
The CSE Index 5-15 years has been developed mainly using data from
the Australian 2006 Census of Population and Housing which collects data
from the whole population. When this paper was written, the 2011 Census
was not available. Future work includes updating the Index using the
2011 Census. Additional data drawn from Year 5 literacy and numeracy
scores from the 2009 National Assessment Program--Literacy and Numeracy
(NAPLAN) provided by the Australian Curriculum, Assessment and Reporting
Authority (ACARA) are also included. We acknowledge that combining data
from different years, i.e. the 2006 Census and the 2009 NAPLAN data is
not ideal, but this is the first year of NAPLAN data that is available
for research purposes. In terms of educational change, we would also not
expect to see much in most communities in the three years from 2006 to
2009.
The Australian Census is conducted once every five years. The 2006
Census data is chosen as the primary data source for this study as the
geographical unit of analysis applied in this index is at the small area
level. The Census is the only source of data available to analyse
multiple dimensions of disadvantage of the population at the Statistical
Local Area (SLA) level, the geographical unit used in this Index.
The SLA which is part of the ABS (Australian Bureau of Statistics)
Australian Standard Geographic Classification (ASGC) is chosen as this
is the smallest unit in the ASGC which does not have issues with
confidentiality and covers the whole of Australia. There were 1426 SLAs
at the time of the 2006 Census. The numeracy and literacy score data are
provided at postcode level, thus a population weighted geographical
concordance is applied to concord these data to the SLA level.
The advantage of using the SLA rather than postcode for our
analysis is that in many urban areas, an SLA is a homogenous suburb,
whereas postcodes are derived by Australia Post for postal delivery, not
data analysis. Postcodes can also be non-contiguous, so there can be a
degree of heterogeneity using postcodes for our analysis, contributing
to the Modifiable Areal Unit Problem, which analysis using the SLA
geography resolves.
However, relying solely on SLAs as a level of geography means
problems can occur in any analysis because SLA boundaries are
administrative boundaries defined by the Australian Bureau of
Statistics, not social or cultural boundaries. Therefore there are a
different number of SLAs in each State, and the number of people in each
SLA in each State is very different. Some small states and territories
have a relatively large number of SLAs and other larger states have very
few. For example, according to the 2006 Census, the Australian Capital
Territory contained only 1.63 per cent of Australia's total
population (including children), but had 7.64 per cent (109) of the
total SLAs. In contrast New South Wales, which contained 33 per cent of
Australia's total population, had only 200 SLAs (or 14.03 per cent
of all SLAs). Queensland also had 479 SLAs (33.59 per cent of total
Australian SLAs), but contained only 19.67 per cent of the total
population. Almost half of Queensland SLAs are Brisbane SLAs, with quite
low populations. The methodology we use to address the issue of uneven
population size is that developed by Baum et al. (2005) and used in Daly
et al. (2008) and McNamara et al. (2009). SLAs in Brisbane and Canberra
(the areas most affected by relatively small population sizes within
SLAs) were aggregated to Local Council Electoral Wards for Brisbane and
Statistical Subdivisions (SSD) for Canberra, so that they were more
similar in population size to SLAs in other areas of Australia. We also
excluded off-shore areas and migratory SLAs from the analysis.
This method was also tested by looking at the Collection District
level variability of the SEIFA score in Tanton et al. (2010), which
found that using the different areas in the ACT and Queensland led to a
similar Collection District variability in each State. This suggests
that the areas used in this analysis are fairly homogenous, which is
important to reduce any aggregation bias or problems with the Modifiable
Areal Unit Problem (MAUP).
The whole process left us with a total of 1 154 small areas for
analysis (after we had aggregated SLAs and removed areas with low
populations and high non-response). The areas are referred as
'small areas' and include SLAs, Wards and SSD to allow
comparison across these areas in Australia.
Domains and Indicators
The first step in the construction of the Child Social Exclusion
index is the choice of dimensions and indicators that need to be
covered. Conceptually, the choice of dimensions and indicators has been
informed by the UN Convention of the Rights of the Child. Bradshaw et
al. (2006: 7) note that--
"the CRC (UN Convention of the Rights of Child) points to the
double role of children as being citizens with entitlements in their own
right and at the same time as being dependent on their families,
schools, communities etc. The discourse on child well-being is thus also
one on child well-becoming. "
The importance of the current wellbeing and future potential of
children --when they become adults (children well-becoming) has
influenced the choice of indicators used to measure social exclusion of
children. Ben-Arieh (2005) emphasises the importance of childhood as a
phase in its own right and the need to get input directly from children
when discussing their wellbeing. The literature also discusses perceived
wellbeing from a child's perspective (see for example, Nic Gabhainn
and Sixsmith, 2005; Wright and Barnes, 2011 and Main and Bradshaw,
2012). For example, when Irish children were asked to identify the most
important determinants of their wellbeing, they emphasised the roles of
family and friends (Nic Gabhainn and Sixsmith, 2005). In a survey of
child material deprivation in the UK reported by Main and Bradshaw
(2012), it was pocket money, saving money and monthly day trips that
were most frequently reported by children as something they lacked but
would like to have. A major international survey of children's
perceptions of their own well-being is currently being undertaken but
Australia is not part of the project (www.childrensworld.org).
In practice, the choice of indicators and domains is also
influenced by data availability. There are indicators which are
important and affect child wellbeing and well-becoming but cannot be
incorporated due to data limitations, i.e. the data are not available in
the Census. For example, crime statistics are important in representing
neighbourhood safety, but are not available for small areas so cannot be
included in this index.
Thus, the selected domains are multidimensional and we combine
indicators that focus on the child's own characteristics, the
child's family characteristics, the child's housing
environment and the child's spatial access to services. All these
indicators cover the five domains considered as important for child
well-being and child well-becoming in Australia. These domains have been
labelled; Socio-economic, Education, Connectedness, Health Services and
Housing. The details of the domains and the indicators are presented in
Table 1.
As the unit of analysis is at the SLA level, the indicators are
mainly provided as the proportion of dependent children aged 5-15 for
each indicator to the total number of dependent children aged 5-15 in
that SLA using the ABS usual residence Census data (so each child who is
not at home on Census night is returned to the area where they usually
reside). Because we are using household level data, there is no
household level information available for households where the children
are not enumerated at home. Effectively, this gives us only children who
were enumerated at home on Census night, as we have excluded all
visitors from the usual residence population.
There are two exceptions to this and these are Year 5 literacy and
numeracy scores and health services. The literacy and numeracy score
results reflect the average scores of Year 5 students in the schools in
an SLA according to the national literacy and numeracy tests and the
ratios of GPs and dentists refer to the number of GPs and dentists to 1
000 population in each SLA. For capital cities, the ratio of GPs or
dentists is at a higher level of aggregation--the Statistical
Subdivision (SSD) rather than SLA is adopted. This adjustment is carried
out as the catchment area for health specialists is likely to be larger
than the SLA.
Cleaning the data included removing SLAs where any indicator had a
'not stated' value (missing data) for all children of more
than 80 per cent, and low cell counts where the total number of children
aged 0-15 years old was less than 30. In addition, the final data also
excludes SLAs where the literacy and numeracy scores were missing for
post codes within these SLAs.
Creating the Index: Methodology
The method used to create the index followed a two-stage approach
(for more detailed discussion on the methodology see Abello et al.,
2012). The first step was to incorporate the indicators into their
domains to create each domain index. The second step combined these
domains into one composite measure/index in which a combination of
principal components analysis and equal weighting techniques was used to
create summary indices. If the indicators were correlated, Principal
Components Analysis (PCA) was used. This is a data summary method which
transforms a set of correlated data into a smaller set of uncorrelated
components that represent all of the information in the original
data-set (see further discussion about this methodology in Salmond and
Crampton, 2002 and Tanton et al., 2010). However, if the indicators are
not correlated, but still considered important to the wellbeing of
children, PCA cannot be used and equal weighting is preferable.
Our correlation matrix shows that most of the indicators are
relatively strongly correlated and PCA is preferred, except for the
housing domain in which low income private renters and overcrowding were
combined using an equal weighting method. The domain indices were
standardised by ranking and then transformed to an exponential
distribution to ensure that the domain weights when they were combined
into a single index were not affected by the different distributions in
each domain (Barnes et al., 2008). Finally an equal weighting method was
applied and a composite measure of the CSE Index at small area level was
constructed.
4. RESULTS
Spatial Distribution of CSE Index 5-15
Figure 1 shows the spatial distribution of the CSE Index 5-15 at
the small area level. The Spearman rank correlation coefficient between
the CSE Index 5-15 and CSE Index for all children aged 0-15 is very
close at 0.94, confirming the close relationship between these two
indices. The 1 154 small areas are divided into four groups based on the
distance from the mean, as follows:
* High risk, if the CSE Index is greater than the mean for all
areas plus 1 standard deviation (SD)
* Moderate risk, if the area's CSE Index falls within one SD
above the mean
* Low risk, if the area's CSE Index is less than or equal to
the mean for the whole sample
* Least risk, if the area's CSE index is less than the mean
for the whole sample minus 1 SD.
Nationally with these classifications, we found that four per cent
of children aged 5-15 years were classified as having a high risk of
child social exclusion and 25 per cent of children in the same age group
were classified as having moderate risk of child social exclusion. As
shown in Figure 1, there are some concentrations of areas with a high
risk of child social exclusion in the capital cities with the exception
of the Canberra. Canberra is estimated as having the most areas with the
least risk of child social exclusion. Note that because of the nature of
area based indicators, this does not mean that there are no children who
are at risk of social exclusion in Canberra (see Goldie et al, 2014 and
Tanton et al, 2015). It means that they are not concentrated in
particular small areas in Canberra. This is a recognised limitation of
area based measures, which is currently being addressed through other
research on individual based indicators of disadvantage (see Tanton et
al., 2015), but is outside the scope of this paper.
[FIGURE 1 OMITTED]
In the middle part of the map (the large black areas), there is a
concentration of the highest risk of social exclusion in the regional
inland areas of the Northern Territory and Western Australia (Figure
1.). Although these areas are large geographically, there are a
relatively small number of children living there.
The disparities in terms of risk of the CSE Index may be clearer
from Figure 2 which shows the proportion of children aged 5-15 years in
each state that fall in to four groups of risk based on the CSE Index
explained earlier. Children in Tasmania and the Northern Territory are
at greater risk of social exclusion relative to other states and
territories. Around 53 per cent of the children in Tasmania were
classified in areas with high or moderate risk of child social exclusion
and 60 per cent in the Northern Territory. In the two states most
affected by the mining boom, Western Australia and Queensland, Western
Australia had one of the lowest proportion of children experiencing a
high or moderate risk of social exclusion, and Queensland had the third
highest of all the states and territories. The conclusion that can be
made is that living in a state experiencing a high level of growth
related to the mining boom did not assure good outcomes for school-aged
children according to this index.
Analysis by Remoteness and Urban/Rural Characteristics
Figure 3 shows the proportion of children that fall in to each of
the four CSE Index risk categories based on the ABS remoteness structure
which classifies locations based on distance to the nearest Urban Centre
or access to various centres of public goods and services (see ABS, 2001
for a description of how these are calculated). The remoteness structure
covers Major Cities, Inner Regional, Outer Regional, Remote and Very
Remote Australia.
While only 4 per cent of children aged 5-15 in the major cities
face the greatest risk of social exclusion, the percentage of children
in the very remote areas of Australia who were in this quintile is 15
times higher (59 per cent) and more than four times higher (17 per cent)
for those who lived in the remote areas. It is interesting also to find
that only one per cent of children aged 5-15 in inner regional and four
per cent in the outer regional respectively face a high risk of social
exclusion. In contrast, only a few of the children who live in either
remote or very remote areas are in the least risk category. It would
appear that the mining boom in remote and very remote Australia has not
in the short term, created a favourable environment for the children
living in those areas.
So, what are the characteristics of children who live in the high
risk areas of the CSE Index. Table 2 shows the proportion of children
living in high risk areas according to their characteristics divided
into urban (capital cities) and rural areas (balance of states). Out of
140 small areas which fell into the high risk areas, 129 of them were
rural areas. Knowledge of the extent and nature of the gap in social
exclusion between urban and rural areas means policy interventions can
be targeted to close this gap. Further, a comparison between individual
characteristics also suggested that some variables may be more important
for urban or rural areas. For example, living in a household with no
motor vehicle may be less important in urban areas where there may be
access to a public transport system that is not available in rural
areas.
Comparison with the Educational Domain
Although child social exclusion is presented as a single
comprehensive measure, further analysis of individual domains that
formed the index and the examination of their relationship with the
index will increase its value to policy makers. One domain of particular
relevance in this study of school-aged children is the education domain.
Thus, 1 154 small areas weighted by the number of children aged 5-15
were also divided into four groups based on the distance from the mean
for this domain. The distribution of children according to the aggregate
CSE index and the educational domain (we refer to it here as educational
disadvantage) is shown in Table 3. There are 65.7 per cent of children
who fall into either low or least risk of CSE and low or least risk of
educational disadvantage or both. At the opposite end of the
distribution, there are 13.1 per cent of children who are classified as
having a high or moderate risk of educational disadvantage or also a
high or moderate risk of child social exclusion or both. While there are
no children that fall into the extremes of high risk on one indicator
and least risk on the other, a small proportion of children, 1.1 per
cent, live in areas that are categorised under high risk of social
exclusion but low risk of educational disadvantage. These areas include
Fairfield-East (Sydney, NSW), Hume-Broadmeadows (Melbourne, VIC),
Biggera Waters-Labrador (rural QLD) and Marngarr (rural NT).
An even smaller proportion of children, 0.1 per cent, lived in
areas that are categorised under high risk of educational disadvantage
but low risk on the CSE index, all in rural areas. These areas were
located in rural Queensland (Bungil, Quilpie, Warroo, Bauhinia,
Isisford, Townsville), rural Western Australia (Corrigin, Mukinbudin,
Carnamah and Port Hedland) and two areas in rural South Australia
(Karoonda East Murray and Unincorp Flinders Ranges).
Comparison with Child Poverty Rates
In this section, the following question is posed: does the CSE
Index 515 provide more information about the spatial distribution of
child disadvantage than the more traditional measure of income poverty?
To address this question, we used child poverty rates in Australia
calculated for children aged 0-14 years (so this will not be a direct
comparison with the CSE Index as the age group is slightly different)
who live in households whose equivalised gross household income is below
the poverty line. The poverty line is set at half-median equivalised
household disposable income which was AUD$ 280 per week or equivalent to
US$ 290 per week (or US$ 42 a day) in 2006. The child poverty rate is
calculated by using a spatial microsimulation model with an extensive
validation process to ensure the reliability of the estimates of child
poverty rates. A fuller discussion of the technique and the validation
process for these estimates is presented in Tanton et al. (2011).
Table 4 provides a transition matrix between the CSE index and
child poverty rates when the rates are grouped using the distance from
the mean following the procedure used for the CSE index. Since there are
some areas that do not have valid child poverty rates, we only include 1
043 small areas in the analysis. The diagonal of this matrix shows an
overlap between the poverty rates and the CSE index and 68 per cent of
children aged 5-15 fall into these categories. It means there are 32 per
cent of children located off-diagonal in the matrix. Further, we also
calculated the Spearman correlation coefficient which shows that both
measures are correlated with R equal to 0.54.
While it is reassuring to observe that no children aged 5-15 fall
into areas with high poverty rates and least risk of CSI Index or visa
versa, Table 4 still shows that 10.9 per cent of children are classified
as living in areas with moderately high poverty rates but low risk
according to the CSE Index, and 4.2 per cent of children are living in
areas with a moderate risk of social exclusion but low poverty rates.
It is worthwhile examining the characteristics of small areas in
the off-diagonal cells. Table 5 shows the characteristics of small areas
which are classified as areas with a moderate risk of social exclusion
but low poverty rates, and compares these with the national average
calculated over the 1 043 small areas. These areas include both urban
and rural SLAs, for example Botany Bay and Marrickville (Sydney, NSW),
Caboolture-East in (rural QLD), Salisbury Bal (Adelaide, SA), George
Town--Pt B (Rural Tasmania) and Alawa (Darwin, NT).
While most of the characteristics are just above or better than the
national average, Table 5 shows that the education indicators,
particularly the outcomes of Year 5 Reading and Numeracy, were slightly
lower than the national average. Access to doctors and dentists was only
marginally above the national average in these areas.
This shows that the CSE index captures different aspects of
disadvantage that are not captured by child poverty rates. This is
important as children in families with incomes just above the poverty
line may still suffer from other types of social exclusion, like
educational exclusion (not being able to afford to purchase some books).
The CSE index, used in quintiles, provides a more nuanced picture of
disadvantage compared to an in poverty/not in poverty indicator based
purely on income.
Children that are not identified as the most disadvantaged by
reference to child poverty rates can now be identified for particular
policy focus using the CSE index. This will be important for regional
planning and for the policy makers to provide better targeted
initiatives.
5. CONCLUSION
This paper has presented the results of an area-based index of risk
of social exclusion for school-aged children. This is an important group
and the focus has enabled further investigation of the education domain
which is of particular relevance to this age group. The results show
that the areas with the highest risk of child social exclusion were in
Tasmania and the Northern Territory and that they were generally outside
the major cities. While Western Australia performed well on the
aggregate index, the other major mining state, Queensland, did not. The
link between state income growth and child disadvantage is complex and
is an important area for future research. The results also show that
school-aged children in remote and very remote areas were at a high risk
of social exclusion.
A further exploration of the education domain shows that
performance in that domain was closely correlated to performance on the
aggregate index. However, there were some small areas where there was a
low risk of social exclusion at the aggregate level and a high or
moderate risk of social exclusion in the educational domain. There were
also some small areas that had a high risk of social exclusion at the
aggregate level but were performing at least at the national average in
the education domain. This highlights the importance of using evidence
from each domain for policy formulation in the relevant area.
The paper also compared a ranking of small areas on the basis of
the CSE index with a ranking based on an income poverty measure. It
showed that while the majority of children fell into the same group as
determined by distance from the mean, there were 32 per cent of children
that were classified differently using the CSE index. The CSE index
captures a more complex measure of disadvantage than income alone.
The aggregate results presented here for school aged children show
a similar ranking of areas to results presented for all children aged
0-15 years reported in Abello et al. (2012) with a rank correlation
coefficient of 0.94. An advantage of focusing on school-aged children is
the ability to investigate more closely some of the factors that are
particularly relevant for this age group.
This paper has demonstrated the advantages of taking a wider
measure of disadvantage than income poverty for identifying small areas
in need of further support for school-aged children. The richer measure
enables policy makers to target resources more effectively to those in
need. The need to raise the performance of schools has been a particular
focus of Australian policy makers in recent years and this CSE index and
its domains offer a useful tool for future policy analysis.
ACKNOWLEDGEMENTS: This paper was funded under ARC Discovery Grant
DP1094318: Towards an enhanced understanding of child and youth social
exclusion risk at a small area level in Australia. The authors would
like to thank the other Chief Investigators and Partner Investigators on
the grant--Prof Laurie Brown, Ms Rebecca Cassells, Prof Asher Ben-Arieh,
Prof Michael Noble, and Ms Leanne Johnson. We would like to thank the
Australian Bureau of Statistics, the Australian Curriculum, Assessment
and Reporting Authority (ACARA), and the Australian Early Childhood
Indicators (AEDI), for supplying data for this project, and the
Electoral Commission of Queensland for providing data to assist in the
aggregation of Statistical Local Areas in Brisbane.
Riyana Miranti
The National Centre for Social and Economic Modelling, Institute
for Governance and Policy Analysis, University of Canberra, Canberra,
A.C.T., 2617, Australia. Email: riyana.miranti@canberra.edu.au
Anne Daly
Faculty of Business, Government & Law, University of Canberra,
Canberra, A.C.T., 2617, Australia.
Robert Tanton
The National Centre for Social and Economic Modelling, Institute
for Governance and Policy Analysis, University of Canberra, Canberra,
A.C.T., 2617, Australia.
REFERENCES
Abello, A., Gong, C., Daly, A. and McNamara, J. (2012). Spatial
Dimensions of Child Social Exclusion Risk: Widening the Scope. Child
Indicators Research, 5(4), pp.685-703.
ABS (2001). Information Paper: ABS Views on Remoteness. ABS Cat #
1244.0, ABS: Canberra
Alkire, S. and Santos, M.E. (2013). A Multidimensional Approach:
Poverty Measurement and Beyond. Social Indicator Research, 112, pp.
239-257.
Annie E Casey Foundation (2010). Kids Count. Online version
accessed, at, http://www.aecf. org/work/kids-count/
Barnes, H., Noble, M., Wright, G. and Dawes, A. (2009). A
Geographical Profile of Child Deprivation in South Africa. Child
Indicators Research, 2(2), pp. 181-199.
Baum, S., O'Connor, K. and Stimson, R. (2005). Fault Lines
Exposed: Advantage and Disadvantage across Australia's Settlement
System. Monash University ePress, Clayton.
Ben-Arieh, A. (2005). Where are the Children? Children's Role
in Measuring and Monitoring Their Well-Being. Social Indicators, 74(3),
pp. 573-596.
Ben-Arieh, A. (2008).The Child Indicators movement: past, present
and future. Child Indicators Research Journal, 1(1), pp. 3-16.
Bradshaw, J., Hoelscher, P. and Richardson, D. (2006). Comparing
Child Well-Being in OECD Countries: Concepts and Methods. UNICEF,
Innocenti Working Paper IWP-2006-03. Innocenti Research Centre,
Florence.
Bradshaw, J., Noble, M., Bloor, K., Huby, M., McLennan, D., Rhodes,
D., Sinclair, I. and Wilkinson, K. (2009). A Child Well-Being Index at
Small Area Level in England. Child Indicators Research, 2(2), pp.
201-219.
Brooks-Gunn, J., Duncan, G.J., Klebanov, P.K. and Sealand, N.
(1993). Do Neighborhoods Influence Child and Adolescent Development?
American Journal of Sociology, 99(2), pp. 353-395.
Burchardt, T., Le Grand, J. and Piachaud, D. (2002). Introduction.
In J. Hills, J.J. Le Grand and D. Piachaud (Eds) Understanding Social
Exclusion, Oxford University Press, New York, pp. 1-12.
Cassells, R., McNamara J. and Wicks, P. (2011).Well-Being Among
Australian Children: a Review of Frameworks and Measures. NATSEM Working
Paper 01/11. NATSEM, University of Canberra, Canberra.
Coulton, C., Korbin, J. and McDonell, J. (2009). Editorial:
Indicators of Child Well-Being in the Context of Small Areas. Child
Indicators Research, 2(2), pp. 109-11.
Daly, A., McNamara, J., Tanton, R., Harding, A. and Yap, M. (2008).
Indicators of Risk of Social Exclusion for Children in Australian
Households: an Analysis by State and Age Group, The Australasian Journal
of Regional Studies, 14(2), pp. 133-154.
Erbstein, N., Hartzog, C. and Geraghty, E. (2013). Putting Youth on
the Map : a Pilot Instrument for Assessing Youth Well-being. Child
Indicators Research, 6(2), pp. 257-280.
Goldie, X., Kakuk, I. and Wood, G. (2014). Two Tales of a City:
Detecting Socio-Economic Disadvantage in an 'Advantaged'
Australian Urban Centre. Australian Geographer, 45(4), pp. 521540.
Goldfeld, W., Katz, I., Tanton, R., Brinkman, S., O'Connor,
E., Mathews, T. and Giles-Corti, B. (2014). Neighbourhood Effects
Influencing Early Childhood Development: Conceptual Model and Trial
Measurement Methodologies from the Kids in Communities Study. Social
Indicators Research, 120(1), PP. 197-212. DOI:
10.1007/s11205-014-0578-x.
Khan, S.A., Brian, D. and Rao, D.S.P. (2000). Child Poverty in
Rural Regions of Australia in 1990. The Australasian Journal of Regional
Studies, 6(1), pp. 47-65.
Land, K. and Crowell, J. (2010). The Child and Youth Well-Being
Index 2010. Online version accessed at, www.soc.duke.edu/~cwi/.
Levitas, R., Pantazis, C., Fahmy, E., Gordon, D., Lloyd, E. and
Patsios, D. (2007). The Multi-Dimensional Analysis of Social Exclusion.
Social Exclusion Task Force: London.
Lupton, R. and Power, A. (2002). Social Exclusion and
Neighbourhoods. In J. Hills, J. Le Grand, D. Piachaud (eds).
Understanding Social Exclusion, Oxford University Press, New York, pp.
118-140.
Main, G. and Bradshaw, J. (2012). A Child Material Deprivation
Index. Child Indicators Research, 5(3), pp. 503-521.
McNamara, J., Tanton, R., Daly, A. and Harding, A. (2009). Spatial
Trends in the Risk of Social Exclusion for Australian Children: Patterns
from 2001 to 2006. Child Indicators Research, 2(2), pp. 155-179.
Miles, R. L., Greer, L., Kraatz, D. and Kinnear, S. (2008).
Measuring Community Wellbeing : A Central Queensland Case Study. The
Australasian Journal of Regional Studies, 14(1), pp. 73-93.
Nic Gabhainn, S. and Sixsmith, J. (2005). Children's
Understanding of Well-Being. The National Children's Office:
Dublin. Online version accessed at,
www.nuigalway.ie/hbsc/.../childrenunderstandingofwellbeing.pdf
Noble, M., Barnes, H., Wright, G. and Roberts, B. (2010). Small
Area Indices of Multiple Deprivation in South Africa. Social Indicator
Research, 95(2), pp. 281-297. DOI 10.1007/s11205-009-9460-7.
Oroyemi, P., Damiolo, G., Barnes, M. and Crosier, T. (2009).
Understanding the Risks of Social Exclusion Across the Life Course:
Families with Children. A National Centre for Social Research (NatCen)
research report for the Social Exclusion Task Force, UK Cabinet Office.
Ramsden, K. (2013). Children's Perspectives on Their Own
Wellbeing: 'I Don't Think They Can Hear Us. Developing
Practice: The Child, Youth and Family Work Journal, 36, pp. 18-30.
Salmond, C. and Crampton, P. (2002). NZDep2001 Index of
Deprivation. Department of Public Health. Wellington School of Medicine
and Health Sciences, Wellington.
Saunders, P., Naidoo, Y. and Griffiths, M. (2008). Towards New
Indicators of Disadvantage: Deprivation and Social Exclusion in
Australia. Australian Journal of Social Issues, 43(2), pp. 175-198.
Sen, A. (1999). Commodities and Capabilities. Oxford University
Press, USA.
Tanton, R., Harding, A., Daly, A., McNamara, J. and Yap, M. (2010).
Australian Children at Risk of Social Exclusion: A Spatial Index for
Gauging Relative Disadvantage. Population, Space and Place, 16(2), pp.
135-150.
Tanton, R., Vidyattama, Y. and Mohanty, I. (2015). Disadvantage in
the Australian Capital Territory. Policy Studies,
DOI:10.1080/01442872.2014.981062.
Tanton, R., Vidyattama, Y., Nepal, B., McNamara, J. (2011). Small
Area Estimation Using A Reweighting Algorithm. Journal of the Royal
Statistical Society: Series A Statistics in Society, 174(4), pp. 931
951.
Wright, G. and Barnes, H. (2011). Defining child poverty in South
Africa using the socially perceived necessities approach. Presentation
at NATSEM Workshop on Child Social Exclusion. 5th July 2011.
Table 1. Child Social Exclusion Index 5-15 years: Domains and
Indicators.
Domains Indicators
Socio-economic In bottom income quintile
No parent in paid work
Single parent family
No family member completing year 12
Education Index of Year 5 literacy and numeracy scores
(compares to the national average)
No internet at home
Connectedness No parent volunteering
No motor vehicle
Health services Ratio of GPs (general practitioners) per 1000
population
Ratio of dentists per 1000 population
Housing High renting and low income (30/40 rules)
Overcrowding (need one or more bedrooms)
Source: Australian Population Census (2006); ACARA (2009).
Table 2. Characteristics of Areas with High Risk of CSE Index 5-15
Years.
Mean Unit National All Urban Rural
average
140 11 129
areas areas areas
Low income % of children 22.6 46.7 43.9 50.2
No parent in paid % of children 15.8 34.5 35.8 32.9
work
Single parent % of children 21.4 32.5 31 34.3
family
No internet % of children 18.5 44 35 54.9
No parent % of children 64.2 78 82.5 72.6
volunteering
No motor vehicle % of children 3.7 17.1 10.2 25.5
High renting % of children 8.4 12.3 14.1 10.1
costs
Overcrowding % of children 8.8 26.7 20.4 34.3
No Year 12 % of children 22.4 42.4 38.2 47.5
Year 5 Reading Mean Year 5 494 440 464 411
Reading
Year 5 Numeracy Mean Year 5 488 448 470 420
Numeracy
GP to 1000 Per 1000 1.5 1.0 1.3 0.8
population persons
Dentist to 1000 Per 1000 0.4 0.2 0.3 0.2
population persons
Note: The national mean is calculated based on the mean values of
1 043 small areas. Source: the Authors.
Table 3. Transition matrix Between the CSE Index and Education
Domain (% children aged 5-15).
CSE vs Education Domain
Education
Domain
Child Social High Moderate Low Least Total
Exclusion risk risk risk risk
Risk
High risk 1.8 1.2 1.1 0.0 4.1
Moderate risk 1.7 8.3 14.6 0.1 24.7
Low risk 0.1 5.4 48.2 3.6 57.2
Least risk 0.0 0.0 3.5 10.4 14.0
Total 3.7 14.9 67.3 14.1 100.0
Source: the Authors.
Table 4. Transition matrix of Child Poverty Rates and CSE Index
(% Children Aged 5-15).
Child CSE Index
Poverty rate
High Moderate Low Least Total
risk risk risk risk
High risk 2.2 5.3 1.2 0.0 8.7
Moderate 1.3 15.2 10.9 0.1 27.4
risk
Low risk 0.0 4.2 37.8 1.0 43.1
Least risk 0.0 0.1 7.8 12.9 20.8
Total 3.5 24.8 57.6 14.0 100.0
Note: The matrix covers 1 043 small areas. Source: the Authors.
Table 5. Characteristics Areas Classified as Low Poverty Rates
but Moderate Risk of CSE Index (% Children Aged 5-15).
Variable Unit Low poverty rates
but moderate risk of
CSE Index
Mean
Low income % of children 23.2
No parent in paid work % of children 18.1
Single parent family % of children 26.8
No internet % of children 23.1
No parent volunteering % of children 68.0
No motor vehicle % of children 5.7
High renting costs % of children 11.4
Overcrowding % of children 11.4
No Year 12 % of children 27.1
Year 5 Reading Mean Year 5 471.2
Reading
Year 5 Numeracy Mean Year 5 466.8
Numeracy
GP to 1 000 population Per 1 000 persons 1.6
Dentist to 1 000 Per 1 000 persons 0.4
population
Variable National Ratio to the
Mean national
mean
Low income 22.6 1.02
No parent in paid work 15.8 1.15
Single parent family 21.4 1.25
No internet 18.5 1.25
No parent volunteering 64.2 1.06
No motor vehicle 3.7 1.54
High renting costs 8.4 1.35
Overcrowding 8.8 1.30
No Year 12 22.4 1.21
Year 5 Reading 494.2 0.95
Year 5 Numeracy 487.9 0.96
GP to 1 000 population 1.5 1.05
Dentist to 1 000 0.4 1.08
population
Note: The national mean is calculated based on the mean values of
1 043 small areas. Source: the Authors.
Figure 2. The Proportion of Children 5-15 Years in State by Risk of
CSE
Index 5-15 (%). Note: The total proportion of children in each
state may not be exactly equal to 100, as a result of rounding to
nearest whole numbers. Source: Author's summary
NSW VIC QLD SA WA TAS NT ACT
High risk 4 1 7 5 2 7 34 0
Moderate risk 26 17 35 29 16 46 26 0
Low risk 55 62 56 48 64 48 39 71
Least risk 15 19 3 17 17 0 2 29
Note: Table made from bar graph.
Figure 3. The Proportion of Children 5-15 Years by Risk of CSE
Index 5-15 and Remoteness Classification (%). Note: The total
proportion of children in each category may not be exactly equal to
100, as a result of rounding to nearest whole numbers.
Major Inner Outer Remote Very
cities regional regional remote
High risk 4 1 3 17 59
Moderate risk 22 26 35 33 16
Low risk 54 71 61 50 23
Least risk 20 3 1 1 2
Source: the Authors' summary.
Note: Table made from bar graph.
