A comprehensive set of child-related monitoring tools: new breakdowns and

3.3 A comprehensive set of child-related monitoring

of children, etc. The breakdowns we put forward for consideration will help in exploring the situation of various subgroups of this age category.

Introducing child age breakdowns

For a selected group of the above indicators of material well-being, a breakdown by child age group (0–5 (0–2, 3–5), 6–11, 12–17) could provide a valuable insight into the situation of children. We consider the at-risk-of-poverty rate, the relative median poverty risk gap, the primary indicator of material deprivation and the housing indicators (both costs and overcrowding) to belong to this group. Since all these indicators are based on EU-SILC, the detail of the breakdown is constrained only by sample size. Our investigations in this respect suggest that the use of a breakdown that differentiates between three stages of childhood (0–5, 6–11 and 12–

17) is fairly safe in terms of statistical robustness for these indicators. However, the introduction of a further breakdown in the lowest age group results in a serious decline in cell size and an increase in statistical uncertainty, which makes it difficult to establish trends for that detail. The option is, therefore, either to increase the sample size of the EU-SILC or to use a less detailed age breakdown, with the concomitant risk of losing some policy-relevant information.

A similar breakdown of the dispersion around the at-risk-of-poverty threshold, of the persistent at-risk-of-poverty rate, and of the secondary indicator of material deprivation may not provide too much value-added. Moreover, a further breakdown of the at-risk-of-poverty rates that are specific to household type, to work intensity and to tenure status would seem to be problematic, given the sample size of the survey. The LFS-based indicator of children (0–17) living in jobless households (an indicator that is an element of the current social inclusion portfolio) could, however, usefully be broken down by child age group.

The need for a new measure of work intensity

The work intensity indicator, as currently developed, relates only to the number of months in the year for which parents (and others of working age in the household) are employed, and overlooks entirely the number of hours for which they are employed during the day or week.⁸⁸ Accordingly, there is a need to take explicit account of part-time working when developing a meaningful indicator of work intensity – not least so that problems of labour-market access and the lack of affordable childcare can be identified. The suggestion is for the new variable to include the effect of part-time working, which will give a more meaningful indication of the attachment of members of the household to the labour market and of their potential access to income from employment. (See Annex 1.2 for a methodological description.)

Breakdown by the new work-intensity variable

The further breakdown of the at-risk-of-poverty rate and the primary indicator of material deprivation among children (0–17) with this new version of the work-intensity variable would facilitate the benchmarking of labour-market and social-care policies in a way that would provide a further insight into policy effectiveness.

Breakdown by household type

In addition to labour-market participation, the composition of households also constitutes a major factor in determining the risk of poverty among children: living with one parent or in a household with two or more other children in itself increases the

88 For example: a single-parent household where the parent concerned works for 8 hours a day every week of the year will be measured as having a work intensity of 1; this will tend (misleadingly) to suggest that, if the household has income below the poverty threshold, the parent can be labelled as among the ‘working poor’ and that the problem therefore has to do with the size of the wage, rather than with working time or access to childcare.

risk of poverty. There are ample grounds, therefore, for the at-risk-of-poverty rate and the primary indicator of material deprivation among children (0–17) to be broken down by household type. However, statistical validation of these breakdowns (on the basis of the available EU-SILC version) shows that the robustness of the estimates is relatively weak in some countries, given the current sample sizes – especially in the case of single-parent households and households with three or more children. On the other hand, the significant variation in the indicators by household type highlights the policy relevance of the breakdown. For these reasons, the suggestion is that the point estimates should be complemented by confidence intervals, in order to aid identification of statistically significant changes over time.

Breakdown by migrant status

To reflect the special situation of children with a migrant background, some of the core indicators could be broken down by the migrant status of the child and/or the parents. The at-risk-of-poverty rate and the primary indicator of material deprivation belong to this group. This report presents empirical evidence (in Section 1.4) of the link between migrant status and the risk of poverty and material deprivation among children (0–17). This analysis also shows that the variable reflecting migrant status might reflect very different social situations (depending on combinations of countries of origin and of destination, and also on combinations of family composition by migrant status). There is, therefore, a case for a more satisfactory measure of this (together with ethnicity).⁸⁹ In this respect, a further addition to the standard EU-SILC questionnaire would be useful. Even a general breakdown between those born in another EU country and those born outside the EU gives rise to robustness problems in a number of countries.⁹⁰ We believe that the EU-SILC might be used as a source to produce illustrative values in certain countries, but in its current design alternative datasets could also be sought to facilitate a European-level monitoring of the situation of migrant children.

A special focus on the Roma

Roma people are especially disadvantaged in the Central and Eastern European new Member States (and in some of the EU-15 countries). Their children are at very high risk of poverty, both in relative and in absolute terms. Also, their material deprivation and housing situation warrant special attention from social policies. To monitor their situation is desirable, though it is clear that the geographical spread is uneven. A special treatment of the complex monitoring of the situation of Roma children is, therefore, suggested, and Annex 3.3 provides a brief summary of this. As a first step, we suggest that breakdowns for Roma ethnic minority status should be provided for most of the commonly agreed indicators (subject to data constraints). A workable definition for the description of Roma status should also be developed.⁹¹

89There are conceptual issues here as well: the current EU-SILC question only explores the stock of migrants, with no information on how long they have been in the country. The ‘non-EU’ category may be far too large and heterogeneous, although sample sizes would also need to be much higher to produce any more detailed breakdown.

90 In most countries, the number of observations – especially in the ‘born in another EU country’ category – is very low, often below 20. Although the number of observations in the non-EU migrant group is somewhat higher, the estimates are still not robust for most countries. Relaxing the definition of migrant status of a family could also be an option here. We defined the migrant status of the family on the basis of the characteristics of both parents, which, in certain circumstances, may prove to be too stringent a precondition.

91 A major source of difficulty is the way in which we measure Roma ethnicity. Simple self-reporting produces serious undercounts, as comparisons of census data and survey data available in Central and Eastern European countries shows (UNDP 2005). A potential solution could be to introduce questions about multiple identities into survey instruments. Also, it could be very useful to have harmonised questionnaire development of the national EU-SILC surveys in those countries where there is a significant proportion of Roma.

The need for new indicators of material child well-being (1) A child-specific material deprivation indicator

While a measure of material deprivation of families with children provides some indication of the overall preconditions of the well-being of children, there are good arguments for identifying a list of items that could be aggregated into a more child-specific deprivation index. For example, having access to educational resources (textbooks, ‘family library’, adequate study place, computer or internet access) clearly constitutes an important material precondition that enables children to perform better at school (and to improve their labour-market prospects in the future).⁹² Differentiation by age group could also be considered.

(2) Share of children in households with low work intensity (as newly defined) An additional and related indicator would be to distinguish households where work intensity (as newly defined to include part-time working) is below a certain threshold – in particular, below 0.5 or some lower value (though greater than 0), which, in most cases, would mean no one of working age in the household in full-time employment (though the appropriate value for the threshold needs to be determined via sensitivity analyses). This, in practice, would cover cases where, for example, a lone parent is employed part time (perhaps because of a lack of access to affordable childcare) or where the only person in work in a couple household is employed part time (perhaps for the same reason). An additional indicator would be where work intensity equals 0.5, which, in most cases, relates to households where only one member of a couple is in employment. This would effectively identify instances where one salary in a couple household is not enough to avert the risk of poverty and material deprivation among the children living in the household. The two possible sources of data to construct the indicator are again the EU-SILC (which is used at present to measure work intensity) and the LFS (which is used to measure jobless households). Details in the LFS on current usual hours of work could potentially be used to construct the indicator (though this would mean ignoring the number of months worked during the year, and so those households in which, say, seasonal working or temporary employment is a source of low income). The EU-SILC has the advantage of including data on both aspects, though it has the disadvantage of having a smaller sample size.

Given the importance of access to employment for household income – and of access to affordable childcare to facilitate this – the indicator could usefully be broken down by child age group (though if the EU-SILC is used as the data source, the size of sample is likely to limit the breakdown).

(3) Childcare (as a service enabling parents to enter labour market)

Given the importance of access to earnings from employment as a means by which households avoid low income and material deprivation, the availability of affordable childcare, which enables parents to work, is a key factor in reducing the risk of poverty among children. Equally, the lack of such childcare is potentially a significant reason for children being at risk. The proportion of children in receipt of formal

92The variables constituting elements of the educational deprivation index developed by the OECD serve as a very useful starting point here. However, the dimension of material deprivation is understood more broadly here than in the OECD typology.

Though from the perspective of the children, the OECD approach of focusing on educational deprivation items only may be very well justified, we work with the general material deprivation items. The reason for this is partly pragmatic (this indicator has just been approved and probed in the Social OMC), but also partly theoretical (material deprivation of the family is a fairly good proxy for educational deprivation as well, while the index of material deprivation may also be a sufficiently good proxy for the general well-being of the household). The indicators on the local environment are grouped with housing in the OECD report (OECD 2009) but considered separately in the EU Task-Force typology. Otherwise, the two (OECD and EU Task-Force) lists of relevant dimensions are the same. The UNICEF typology has a broader coverage for relationships, and considers subjective well-being of children as a separate dimension, while it includes fewer details of material well-being (UNICEF 2007). See details in OECD (2009).

childcare, as calculated from the data on this in the EU-SILC, is, accordingly, an important indicator of the access of parents to employment and the income from this.⁹³ (Formal care is defined as covering education at pre-school or compulsory school and care at centre-based services and day-care centres; it is distinguished from informal care, such as with a childminder, friend or relative, because of the specialised nature of the care provided, which tends to make parents more willing to leave their children in such places.)

In view of the varying importance of access to childcare at different ages, the indicator needs to be broken down by child age group, preferably in line (at least approximately) with the Barcelona targets, which specify the proportion of children aged under 3 and the proportion aged 3 up to primary-school age to be covered by childcare. Arguably, however, it is also relevant to monitor the availability of childcare outside school hours for children of primary-school age up to the age of 12 (around which time children can look after themselves), since this might determine whether it is possible for parents to be employed full time, or whether at least one of them can work only part time because of the need to look after their child. For the same reason, the hours of childcare received are equally relevant for children in younger age groups. A supplementary indicator, therefore, for the age groups 0–2 and 3–5 could be the proportion of children in each group who receive formal care for 35 hours a week or more. Ideally, the indicator should also be broken down by household type, distinguishing especially lone-parent from couple households, since the availability of childcare is a particularly important determinant of the ability of single parents to take up employment. The small sample size of the EU-SILC, however, makes such a breakdown problematic, given the need also to divide children up by age group.

Childcare services are, however, important as an end in themselves. Childcare provides an opportunity for children to develop better social skills in pre-school institutions and to benefit from care from professional personnel in formal and less formal (but socially organised) care institutions. Therefore, in addition to the availability of childcare services, both the quality and accessibility of childcare are important aspects that contribute to child well-being, insofar as a lack of quality and lack of accessibility deter parents from using the facilities. Therefore, there is a need to monitor quality frameworks for childcare, so that Member States are encouraged to collect and report service-quality measures. This may help resolve the potential conflict between the interests of children and those of parents – especially, but by no means entirely, at an early stage in the development of children.

Non-material indicators: (1) education, health and exposure to risk-taking behaviour

Selection of commonly agreed non-material indicators that are relevant for the situation of children

An overview of the commonly agreed health and education indicators (covered in the Social OMC or the Education OMC process) shows that the further refinement of a set of these indicators could also help to monitor the situation of children across the EU. The indicators monitored at present are:

• Early school-leavers: share of persons aged 18–24 who have only lower secondary education and are no longer in receipt of education or training (B1).

93 A preferable indicator is arguably the proportion of households with children in which the youngest child is in receipt of formal childcare, since this focuses on the household rather than on the child as such, and since arranging childcare for the youngest child is typically the key constraint on the ability of parents to work. This, however, is more complicated to calculate than the simple proportion of children in a particular age group receiving childcare, even if the latter is a potentially misleading indicator of the households with access to childcare, since it is affected by variations in the number of children per household. Such variations, however, tend to have a relatively small distorting effect on the measure.

• Low reading literacy performance of pupils aged 15 (B1).

• Infant mortality (B2).

• Infant mortality and breakdown by socio-economic status (NAT) (B2).

• Perinatal mortality (NAT) (B2).

• Vaccination in children (B2).

• Life expectancy at birth (B2).

• Life expectancy at birth by socio-economic status (B2).

Although the definition of all these indicators has already been agreed, some further elaboration might be useful in order to capture the broadest possible range of aspects in relation to the three dimensions (education, health, exposure to risk and risk behaviour) in question.

Education outcomes: what (else) to monitor?

There are two basic types of indicator to monitor child well-being in terms of educational outcomes: on the one hand, participation in education or training from pre-school to upper-secondary education and beyond; on the other, the performance of children in assimilating what they are taught. With regard to participation, this is (though only at a late stage) partially covered by the early school-leaver indicator, though it could be supplemented by an indication of attendance at pre-school at the other end of the scale (receipt of childcare by those aged 3–5 picks this up to some extent).

With regard to performance, various surveys differentiate between reading literacy, mathematical literacy or numeracy, and basic understanding of science literacy. The starting point in this respect is that a measure of low reading literacy of pupils aged 15 is already part of the agreed portfolio. There are, however, good arguments for monitoring literacy performance at an earlier age, especially since there are widely available data sources for children at 10 years of age.⁹⁴ A particular issue here is which of the three literacy indicators should be used in monitoring. While there are good reasons to combine them (as occurs in the OECD report),⁹⁵ there is also a case for keeping them separate and for focusing on reading literacy. Though the difference may not be large, in terms of equity, reading literacy may be more relevant than the other two (especially when ethnicity and migrant status are considered), since it measures competencies that are more basic than the other two (from the perspective of social inclusion).⁹⁶ The main suggestion here is to supplement the present PISA- based indicator with an indicator of children’s reading literacy at age 10, based on PIRLS,⁹⁷ broken down in various ways.

Health outcomes: what (else) to monitor?

A number of already agreed indicators in the health portfolio of the Social OMC serve to characterise well the broad aspects of child health across the EU. Among the present indicators, however, infant mortality and perinatal mortality tend to refer to child health in a ‘negative’ way.⁹⁸ There is arguably a need, therefore, to make a better assessment of the health status of children at various ages. In Table 3.2 (and

94 Like PIRLS (Progress in International Reading Literacy Study).

95 OECD 2009: 40.

96 Another argument for keeping these separate is that science and maths literacy scores for 10-year-olds are collected in surveys (TIMSS), separate from the reading literacy scores for the same ages in PIRLS.

97 See Table 3.2, as well as and Annexes 3.2 and 3.5.

98In addition to this, a methodological development on avoidable child mortality would be very important, even if there is a data gap in terms of comparable information on various reasons for avoidable mortality, while the level of development of the health systems also renders it very difficult to justify what deaths are ‘avoidable’.

in Annexes 3.2 and 3.5), indicators of low birth weight,⁹⁹ breastfeeding of babies,¹⁰⁰ and measures of obesity of those aged 11–15 are suggested, along with some measures of health behaviour, such as eating breakfast and fruit daily and doing physical exercise regularly. An overall measure of self-perceived general health is also put forward for consideration.¹⁰¹ The relevance of these indicators is supported by research, which also points to their responsiveness to policy.¹⁰²

Equity considerations reflected in health and education outcomes: new breakdowns for some already agreed ‘child outcome’ indicators

Some child outcome indicators of health or education performance do not fully capture the equity aspect of child well-being. To measure and monitor this, the introduction of parental background is important. This is reflected in the fact that in the health portfolio it is already agreed that life expectancy and infant mortality indicators need to be broken down by parental socio-economic status.¹⁰³ By the same reasoning, it is arguable that both the reading literacy performance indicator of 15-year-old children and the newly suggested indicator at age 10 could usefully be broken down by education of parents.¹⁰⁴ It is equally arguable that they should also be broken down by the migrant status of parents.¹⁰⁵ For the other newly suggested indicators, further breakdowns do not seem necessary.¹⁰⁶

Risk-taking behaviour and exposure to risk: what to monitor?

There are differences of views as to what behaviour constitutes a long-term risk for children. However, there seems to be a consensus that alcohol, smoking, drug use

99 A large corpus of literature has built up showing that low birth weight has far-reaching negative effects on future educational attainments (Currie et al. 2008; Currie and Hyson 1999; Johnson and Schoeni 2007). Another study finds that childhood health characteristics have a significant impact on earnings among men at age 33 (Case et al. 2003). These effects appear to operate through the effect of poor childhood health on educational attainment, early adult health and initial earnings. However, relatively little knowledge is available on how child health at birth affects schooling outcomes: whether it matters primarily because it predicts future health or through some other channels (Currie et al. 2008).

100 Although the data situation for breastfeeding is very poor. See on this Indicator table B2.6 in Annex 3.5.

101 Most of the health behaviour variables are available for ages 11, 13 and 15 in HBSC. We propose to cover the earliest possible ages. This, however, should be decided bearing in mind that there is always a compromise between the need to cover early ages and the meaningful expectation of having reliable data from interviews with children themselves.

102 Research indicates that post-natal parental investments can compensate for low birth weight. It is found that i) mothers are more likely to delay kindergarten entrance for their low birth weight (LBW) children; ii) birth weight impacts labour-force supply following birth among women who worked prior to birth. Delaying kindergarten entrance by one year substantially increases both maths scores and reading scores, but curtailing maternal labour supply following birth has little impact on test scores. Little evidence is found that LBW children benefit differently from delayed kindergarten entrance or curtailed maternal labour supply.

LBW children, however, benefit differently from family size than do their normal birth weight peers. The adverse effects of LBW are substantially diminished for children raised in smaller families (Loughran et al.2004). Research also suggests that public programmes that focus on the post-natal period can help LBW children to catch up with their normal birth weight peers (Loughran et al. 2004).

103 It is understood that a taskforce set up by Eurostat is working on the difficult job of developing this breakdown.

104 Our line of reasoning is simple here: in both PISA and PIRLS there are attempts to develop other indicators of socio-economic status of parents. The methodologies and background concepts are very different, but education is a good proxy for both socio-economic indices and it is easier to make the education status comparable across surveys and countries.

105 Where data are available, pupils with migrant background have lower scores, though this depends to various extents on country experiences and on the origin of migration into the countries in question. Also due care to the interpretation of the results is warranted (Song and Róbert n.d.).

106 No strong evidence is found that the effects of infant health on educational outcome are any worse among families with low socio-economic status (SES) (i.e. families from the bottom two residential income quintiles) (Oreopoulos et al. 2006; Black et al.

2005). Indeed, a study using British data (NCDS) indicates that high SES boys suffer more from LBW than do their low SES peers, in terms of poorer educational attainment (Currie and Hyson 1999). Results suggest that effective interventions to combat the effects of LBW are important in all sectors of the population (Currie and Hyson 1999), though poor children are supposed to suffer from double jeopardy, in that they are both more likely to suffer negative shocks and less likely to be able to recover from them (e.g. Bradley et al. 1994). Currie and Hyson (1999), however, find that low birth weight has adverse effects on children generally, irrespective of whether they have high or low socio-economic status. No significant differences are found in the effects of birth weight by mother’s education, by family income, or by birth order of the children (though this might be due to the small sample size) (Black et al. 2005). Results highlight the importance of finding effective interventions to combat the effects of low birth weight for both high and low SES children (Currie and Hyson 1999). Nor did we find support for parental SES breakdowns for the other suggested health indicators (breastfeeding, obesity or the health behaviour variables).

and early sexual experience (more importantly, child-age pregnancy) can affect the well-being and well-becoming of children in both the short and the longer term.¹⁰⁷ The choice of indicators in this area is also very much dependent on data availability. To date, there are no agreed indicators in the Social OMC of risk-taking behaviour. In Table 3.2 (and, for details, in Annexes 3.2 and 3.5), indicators of teenage births,¹⁰⁸ smoking, alcohol consumption and drug use are suggested for consideration (it should be noted that, for some of these indicators, the evidence is divided as to their effects on educational and other outcomes at a later stage of life).¹⁰⁹ The three latter indicators of risk-taking behaviour can be measured separately, but there are good reasons to suggest that they be monitored and presented together: alcohol, drugs and smoking tend to be correlated to some extent, and so measuring all three together is advisable.¹¹⁰ Participation in criminal activity and being a victim of crime are also potentially important aspects of risk-taking behaviour, in terms of both present well-being and future outcomes. However, data availability and cross-country comparability at present limit viable suggestions, and consideration of how to improve things in this regard would be useful.

Measurement of risk-taking behaviour

Measuring the use of alcohol and various substances is difficult. Regulations differ from country to country, different wording may cause major discrepancies in survey results, and the sampling frames of surveys can vary. Data on alcohol and drug use are also sensitive to the age of respondents,¹¹¹ while there is an issue over whether questions should refer to lifetime use or use in the last week or month, as well as over the frequency of use.¹¹² Similar difficulties arise as regards other types of risk-taking behaviour. The number of teenage births may be very different from that of teenage

107 The UNICEF report card lists smoking, drunkenness, cannabis use (all averaged for three ages: 11, 13 and 15) and having had sexual intercourse (for the 15-year-olds), teenage births (15–19 women) and some experience of school violence (fighting in the previous 12 months, being bullied in the previous 12 months – both averaged for three ages: 11, 13 and 15) (UNICEF 2007: 30–31). The OECD report focuses on regular smoking among 15-year-olds, having been drunk (13- and 15-year-olds) and teenage birth rates (women aged 15–19) (OECD 2009: 52–57). Also, the penetration of bullying at school is presented as a form of risk in terms of the psychological and social development of children.

108 First-generation studies on the topic concluded that teen childbearing had a strong negative effect on mothers in terms of low levels of education, employment and earnings, and high levels of dependence on welfare, etc. More recent studies (with more reliance on multivariate analytic tools) have revealed that some of the differences in well-being between adolescent and older mothers are most likely due to factors other than teenage motherhood in itself. Research indicates that teen mothers come from more disadvantaged backgrounds than do women who delay childbearing. Teen mothers grow up in poorer homes and with less well-educated parents than women who do not have a child as a teenager. They are also more likely to grow up in single-parent families than are women who delay childbearing (Hotz et al. 1997), but see also Geronimus and Korenman (1993) and, more recently, Fletcher and Wolfe (2008).

109 The results on the educational effects of teenage alcohol use are somewhat controversial. Research based on US data reveals that drinking onset has, at most, only a modest effect on educational attainment (Koch and Ribar 2001). In the US, heavy (binge) drinking is found to reduce school performance (DeSimone and Wolaver 2005) and the probability of receiving a high-school diploma, and to increase the probability of graduating with a General Educational Development (GED) diploma (Renna 2008). A British study reports that heavy drinking in adolescence has a negative effect on post-secondary degree completion by age 42 among males (independent of childhood risk factors correlated with both heavy drinking and school achievement). The effects of teenage alcohol use, however, vary depending on the social background of the individuals: heavy alcohol use appears to be more hazardous for working-class males than for males from more advantaged backgrounds (Staff et al. 2008). Though, as teen alcohol use lowers academic performance or educational attainment, its potential impact on labour-market outcomes during adulthood is little researched.

Similarly, the effect of teen drug use is to reduce educational attainment (though composition effects suggest we should exercise caution here, too). Based on data from the US, a recent study shows that marijuana consumption has a significant negative short-term effect on schooling performance (Pacula et al. 2003).

110 Arguments in favour of a joint measurement and a collapsing of the information content of these measures on individual level into a single index can – in principle – also be proposed. However, for an aggregate index like this, further studies are necessary, with input from experts of risk behaviour.

111 When comparing ESPAD and HBSC survey results, the most recent ESPAD report suggests that penetration data from two different surveys can be compared when the difference in the mean age of respondents is no more than plus or minus 0.2 years (see the ESPAD 2007 report: under Hibell et al, 2009). Some of these comparisons are also quoted in greater detail in Annex 3.5, indicators B3.2–B3.4.

112 In addition, the ‘severity’ of drinking, gauged by a variety of measures of consuming large amounts of alcohol in a relatively short period, is also an issue for methodological discussion. While the definition of ‘heavy’ or ‘binge’ drinking may be problematic as well, the definition of ‘being drunken’ is much more culturally framed (see Plant and Plant 2006; Elekes 2007).

In document Child poverty and child well-being in the European Union (Pldal 187-197)