Out of Sight, Out of Mind: Ethnic Inequalities in Child Protection and Out-of-Home Care Intervention Rates

Abstract

This paper examines the interlocking roles of ethnicity and deprivation in producing inequities in the proportion of children who are subject to state child protection interventions. In contrast to the USA, ethnic inequities have had little attention in research or policy in the UK and across Europe, and administrative data are limited and methodologically weak. A study of over 10 per cent of all children on child protection plans or who were looked after in out-of-home care in England in March 2012 is reported. Children from ethnic minority categories were much more likely than ‘White’ children to be living in disadvantaged areas and this has to be taken into account when examining intervention rates. Controlling for deprivation and examining small subgroups of the broad ethnic categories radically alters the simple understanding that ‘Black’ children are overrepresented compared to White amongst children in out-of-home care, while ‘Asian’ children are under-represented. While this study could not explain these patterns, it reinforces the importance of both socio-economic circumstances and ethnicity for understanding inequities in intervention rates. The evidence underlines the powerful moral and economic case for action to reduce inequities in powerful state interventions in family life, not only in England, but internationally.

Introduction

Child welfare inequality is defined as occurring ‘when children and/or their parents face unequal chances, experiences or outcomes of involvement with child welfare services that are systematically associated with structural social dis/advantage and are unjust and avoidable’ (Bywaters et al., 2015, p. 100). Fundamentally, the case for greater equity in children’s social care rests on arguments about social justice and the state’s role in protecting family life (Bywaters et al., 2015). Additionally, one might expect large variations in the proportion of children receiving different kinds of social care interventions (‘intervention rates’) to be a significant focus of attention on grounds of cost. Two key variables affecting intervention rates have been identified in previous reports of a study of child protection services: family socio-economic circumstances and ethnicity (Bywaters et al., 2014a, 2014b). The interaction between these factors has received only limited attention in the UK or Europe in recent years.

In 2014, the Department for Education (DfE) in England published two papers on children’s social care outlining the ‘research priorities and questions’ which should inform the development of policy and practice (DfE, 2014a, 2014b). In neither paper is there any mention of race or ethnicity as a subject requiring study, despite clear evidence of large differences in intervention rates between ethnic categories. Owen and Statham (2009) found substantial levels of ‘disproportionality’ in child protection and out-of-home care populations between broad ethnic categories. Moreover, there is remarkably little recent published research on these differences in the UK or in Europe more widely. Perhaps it is because it has ‘long been known that Black and Mixed ethnicity children are over-represented within the Children Looked After population (relative to their numbers in the overall population) and that Asian children are under-represented’ (Owen and Statham, 2009, p. 6). The association between ethnicity and intervention rates has, perhaps, become—for many—a taken-for-granted backdrop to practice, requiring neither interrogation nor action.

The handful of UK papers on race and child welfare published recently have included several reviews of earlier research (Chand, 2005; Chand and Thoburn, 2005, 2006; Barn, 2007) and empirical studies of social workers’ responses to ethnicity (Williams and Soydan, 2005); Black African children’s experiences of the child protection system (Bernard and Gupta, 2008) and pathways to adoption (Selwyn and Wijedesa, 2011). While valuable, only Owen and Statham’s study has attempted to quantify current inequalities in key intervention rates between ethnic categories and to understand the role of deprivation. Moreover, Owen and Statham’s work, based on 2004–06, was limited in two major ways. First, they relied on official data employing only five broad ethnic categories: White, Mixed, Asian, Black, other; and, second, they were unable to control for family or neighbourhood deprivation as no such data are routinely collected by central government in England.

This, of course, contrasts strongly with the research record in North America, where extensive work has focused on issues relating ethnicity to child protection and out-of-home care, and on disproportionality and disparity (e.g. Hill, 2006; Putnam-Hornstein et al., 2013; Klein and Merritt, 2014; Maguire-Jack et al., 2015). At risk of over-simplifying, a small number of important broad conclusions appear to emerge. First, Black children are overrepresented in the US child protection system compared to White children, using the crude tripartite categories of Black, White and Hispanic (Drake et al., 2011). This is reported to result not primarily from differential treatment by services, but from greater exposure to damaging socio-economic conditions. Second, however, despite similarly disadvantaged material and environmental circumstances as Black populations, children of Hispanic backgrounds are under-represented. It is suggested that this is because of protective cultural factors with a strong value placed on the family, sometimes without proper recognition of the impact of the history of slavery, segregation and racial discrimination on Black family life. Third, even after poverty is taken into account, there remains a gap in the proportion of White and Black children who are placed in out-of-home care. ‘(B)lack placements are, on average, higher than those for whites but less elastic with respect to the level of Black social disadvantage measured at the county level’ (Wulczyn et al., 2013, p. 73). The social gradient is less steep for Black children. Each incremental increase in disadvantage produces a smaller increase in placement rates for Black children than for White. Of course, such findings cannot be simply transferred to the UK. However, the central focus of US research on equity is instructive and the research methods and theoretical developments worthy of consideration.

To understand inequalities in intervention rates, we have proposed a model of intervention chances—the proportion of children in a given area or group who are in receipt of a particular kind of intervention—which requires empirical testing (Bywaters et al., 2015, p. 104). Two broad interlocking forces are at work: demand and supply (see Supplementary material, Figure 1). For example, the demand for child protection interventions, reflecting the level of maltreatment in an area, will be affected by factors influencing the capacity and capabilities of families and communities to parent effectively. Such factors include the economic, environmental and cultural context, including the impact of institutionalised discrimination. These elements, including the consequences of housing and welfare benefits policies, impact in varied ways on family life in all geographical areas. In disadvantaged areas, pressures may contribute to high levels of neighbourhood violence and conflict with weakened social bonds, although in some circumstances they may generate greater social solidarity and the creation of alternative networks of support (Maguire-Jack and Wang, 2016). Geographical areas where more parents can buy a range of services and resources to support parenting tasks are likely to be areas of relatively low demand.

However, intervention rates are also affected by the supply of services. Areas with few services, where services are not accessible or appropriate to families’ needs, or which cannot reach some communities are likely to have lower rates of intervention than areas where services are plentiful and accessible. (By services, we do not necessarily mean interventions that are welcomed by families.)

Of course, both supply and demand factors contain complex and contradictory elements. Deprivation alone is not a sufficient account of family pressures. Individuals vary in their resilience as a result of a range of historical factors, including those affected by disadvantage (Bramley and Fitzpatrick, 2016). Equally, while the distribution of resources for service provision may take demand into account, less pressurised areas may attract a more stable, experienced and qualified staff group, thus acting against the distribution model. Similarly, the underlying cultures of service providers may affect performance and outcomes. For example, certain socio-economic or ethnic groups may either be stigmatised, leading to a higher chance of intervention, or staff may become desensitised to need in high deprivation areas, leading to a lower chance.

While official statistics covering children on child protection plans (CPP) or who are looked after (LAC) in England include some analysis by ethnic category, they do so only in a restricted way (DfE, 2015a, 2015b). A number of limitations can be identified in addition to the lack of data about families’ material circumstances. First, although data are collected on all ‘children in need’ (including CPP and LAC) using eighteen ethnic categories (Supplementary material, Table S1), published reports on CPP only use the five broad categories outlined above. Some reports on LAC use all eighteen. Second, no CPP or LAC rates are published by ethnic categories. Third, in 2015, data are only given about ethnicity in national tables, although local-authority-level data were available in 2014 and earlier.

Moreover, there are fundamental problems in the application of ethnic categories. The guidance on statistical returns gives no indication how they should be collected. Who decides what category is recorded? How likely is it that parents, young people or social workers will interpret these categories in the same way? Many people with dual heritage may identify themselves as Black, as a category defined by racism rather than biology. Others may view their ethnic identity as ‘African’ despite having been born in the Caribbean or the UK. The ‘Mixed’ category poses further challenges, as the gender of the parent in a given category is not included. This may mask differential intervention rates between Mixed-heritage children raised by both parents, those raised by a single White parent and those raised by a single Black or Asian parent.

The eighteen-category approach does not solve either these empirical or philosophical problems. It still involves very broad groupings, such as merging all Black British people of African heritage or all Asian British Indians into single entities. It still involves a kaleidoscope of categorical systems based on colour, continent, country or other basis of difference. For this reason, throughout this report, whenever appropriate, we use the term ethnic category rather than ethnic group to make it clear that the categorisation is externally applied rather than a matter of a chosen identity involving community ties. While we describe children as being in a particular category, ‘White’ or ‘White British’, for example, we are aware that these are fluid and contested descriptions. What is required is a clearer theoretical rationale for the categories chosen, greater attention to ensuring consistency in the application of categories in official data and a greater willingness to report what is found.

Method

Following ethnical approval from Coventry University and the Association of Directors of Children’s Services Research Group, thirteen local authorities (LAs) in the English West Midlands region provided data on all children who were either on a child protection plan or in out-of-home care on 31 March 2012, the census date for annual returns by LAs to the Department for Education which are the basis for official statistics. The LAs, covering urban and rural areas, were responsible for nearly 1.2 million children aged from birth to seventeen, 10.5 per cent of all children in England, 10.6 per cent of all children on a CPP and 11.3 per cent of all LAC on the census date. The data included each child’s age, gender and ethnic category, using the eighteen prescribed categories outlined in the Supplementary material, Table S1. In addition, as a proxy measure of family socio-economic circumstances, LAs identified the child’s local neighbourhood or, for LAC, the home address from which they entered out-of-home care. The neighbourhoods, known as ‘lower super output areas’ (LSOAs), covering an average of 1,500 residents, are an element of the national structure of geographies on which official statistics are based. Because of low numbers in some ethnic categories, data were grouped into middle-layer super outputs areas (MSOAs; n = 696)—combinations of four or five contiguous LSOAs with an average population of 7,200.

We analysed the relationships between rates of intervention and deprivation using age-based population counts from the 2011 Census and 2010 Index of Multiple Deprivation scores (IMD). The IMD is a broad measure of deprivation encompassing seven key dimensions and thirty-eight indicators. To estimate MSOA deprivation ranks, a population weighted average of LSOA scores was calculated for every MSOA in England. These were then divided into deciles or quintiles ranked in terms of IMD and the MSOAs in our sample located accordingly. In subsequent tables and charts, findings for quintile 1 refer to all those neighbourhoods (MSOAs) in the sample which were amongst the 20 per cent least deprived neighbourhoods nationally. Quintile 5 refers to those neighbourhoods which were amongst the 20 per cent most deprived neighbourhoods nationally. The study methods are described in more detail in an earlier paper (Bywaters et al., 2014a). In reporting this analysis, we sometimes analyse the data in terms of the broad ethnic categories to support comparisons with earlier studies. We sometimes omit particular categories when they contain too few children for the analysis to be reliable.

It must be remembered that this is not a study of the relationship of the circumstances of individual families to intervention rates because data on family circumstances are not available either in official data or in other research. We use the deprivation levels of small neighbourhoods as a proxy for family circumstances but cannot know whether the families whose children receive children’s services interventions mirror other families in the neighbourhood. This is a particularly important caveat for the data on ethnic minority groups. For example, it is possible that structural impediments to housing mobility and other impacts of racism, and the attraction of remaining close to family members, communities with which they have strong cultural ties and schools in which their children are not isolated may mean that minority families are more likely to remain in poorer neighbourhoods even when their own financial circumstances improve. However, this study could not provide evidence to confirm or deny this conjecture and other factors may also be at work.

Findings

Demography and deprivation

In order to contextualise the data about child welfare interventions, it is necessary to explore, first, the distribution of children from different ethnic categories between and within LAs and, second, the intersection between ethnicity and deprivation. Overall, in the West Midlands, just over 70 per cent of all children were identified as White in the 2011 Census, with 16 per cent Asian, 5 per cent Black and 6 per cent as of Mixed heritage. However, this picture masks extreme differences between LAs, with White children accounting for 97 per cent of children in Herefordshire but only a little over 40 per cent of children in Birmingham (Supplementary material, Table S2). This diverse and changing ethnic profile of England’s children underlines the need for more refined data and an enhanced focus on ethnic disparities.

The proportion of children in the population identified as White has reduced over recent years. Amongst children aged from birth to four, fewer than seven in ten were White compared to over three-quarters of sixteen- to seventeen-year-olds (Supplementary material, Table S3). For all non-White ethnic categories, the proportion under five years is greater than for the over-fifteens. As we shall demonstrate later, because of large ethnic and age diversities in the rates of children’s services interventions, these differences in the child population between LAs and over time have a significant relationship with demands on the services.

Of course, these patterns, because they only focus on broad groupings, mask as well as reveal differences in the populations of children that different LAs are serving. For example, as Table 1 shows, the distribution of ‘Asian’ children between those of Indian, Pakistani, Bangladeshi and other origins varied widely. Birmingham had four times as many children from a Pakistani background as from an Indian background, while, for Wolverhampton, the proportions were roughly reversed. In Sandwell, there were 50 per cent more ‘Black’ children from a Caribbean background than an African background but, in Coventry, there were eight times more African than Caribbean children. Again, such differences have a crucial impact on how overall intervention rates should be interpreted.

A second major consideration is the intersection between ethnicity and deprivation. According to the 2011 Census, over 38 per cent of all West Midlands children were living in the most disadvantaged 20 per cent of neighbourhoods nationally—almost twice the rate that would occur by chance—while only 12 per cent lived in the most advantaged 20 per cent. In every quintile except the most disadvantaged, children were under-represented. However, this varies strikingly between LAs and by ethnic category. One way to show how deprivation interacts with the distribution of children is to sort the thirteen LAs into three groups (Table 2): those in the most disadvantaged third of LAs in England (six LAs), those in the middle (two LAs) and those in the most advantaged third (five LAs). In the most advantaged third, over 50 per cent of children lived in the 40 per cent most affluent neighbourhoods nationally, compared to only 8 per cent in the disadvantaged third. In the most disadvantaged third of sample LAs, 64 per cent of children lived in the most deprived 20 per cent of areas, compared to just 10 per cent of children in the most advantaged third of LAs. As the chances of a child being on a CPP in quintile 5 is six times greater than in quintile 1, and the chances of being a LAC is five times greater, the impact of these demographic patterns on services is very considerable, with powerful implications for wider social policy and future life chances.

Further stark differences are revealed between ethnic categories (Table 3). Overall, two-thirds of all Asian children and more than three-quarters of all Black children were living in the most disadvantaged 20 per cent of neighbourhoods in the country, but only a little over a quarter of White children were. However, in the most disadvantaged third of LAs, these proportions increased to almost three-quarters of all Asian children and more than four-fifths of all Black children. This is an extraordinary concentration of children from these broad minority ethnic categories in the most disadvantaged neighbourhoods—markedly different from the pattern for White children. Again, drilling down reveals further differences, particularly between Asian children of Indian background and those with Pakistani or Bangladeshi or other Asian backgrounds. Forty-six per cent of children of an ‘Asian Indian’ background lived in quintile 5 compared to almost 80 per cent of ‘Asian Pakistanis’ and almost 85 per cent of ‘Asian Bangladeshis’ (Supplementary material, Table S4). However, the proportion of children of an Indian background living in the most deprived 20 per cent of neighbourhoods is still approaching double that of White British children.

Previous analysis suggests that the socio-economic inequality may be more acute in the North and Midlands than in London and the South of England, with African, Bangladeshi, Other Black, Pakistani and Caribbean groups being most disadvantaged (Jivraj and Khan, 2013). Our sample may not be representative of the country as a whole. However, this is a 10 per cent sample of all children in England, not a small sub-group.

Deprivation and intervention rates: broad ethnic categories

Implicitly taking rates for White children as the norm, the main recent UK study of ethnic differences in intervention rates (Owen and Statham, 2009) concluded that children of Mixed heritage were overrepresented amongst CPP and LAC, Asian children were under-represented and Black children were overrepresented amongst LAC but under-represented amongst CPP. At the whole LA level, the rates for our sample (Table 4) show a similar pattern, with rates for Mixed heritage children the highest and for Asian children the lowest across both CPP and LAC. Asian rates were about 50 per cent of White for CPP but only 25 per cent for LAC. Interestingly, for White, Mixed and Black children, LAC rates were much higher than CPP rates but, for Asian children, they were lower. Rates for Black children in our sample were around 10 per cent higher than for White for LAC but around 20 per cent lower for CPP.

However, this overall view gives only a partial perspective because it fails to take into account the patterns of deprivation outlined earlier: it fails to compare like with like. After controlling for deprivation by examining rates in quintile 5 where most Black and Asian children lived, both Black and Asian children much less likely than White children to be on CPPs or to be LAC.

Table 5 shows that, in quintile 5, where more than half the Mixed heritage children lived, they had the highest CPP rates, but the gap between White and Mixed heritage children is greatly narrowed. When comparing like with like, there was little difference. The overall much lower White children’s rate was a reflection of the population distribution, with a far larger proportion of White children living in more advantaged neighbourhoods. When comparing Q5 only, Asian children’s CPP rate was around a third that for White children. The gap in rates between Black and White children also increased, from less than 20 per cent to over 55 per cent.

Similarly, Table 6 shows that, unless controlled for deprivation, a partial impression is given of relative LAC rates. In Q5, The gap between White and Mixed heritage children was much reduced, from nearly double (Table 4) to under 30 per cent. Black children, far from being overrepresented compared to White, were more than a third less likely to be looked after. Asian children were six times less likely than White children to be LAC. If the Asian rates had been applied to White children, there would have been around 500 looked after White children in Q5 in our sample LAs rather than nearly 3,000.

Of course, none of these arguments means that the social justice case for action on inequities in intervention rates is no longer relevant. The more detailed data controlled for deprivation do not smooth away inequities in intervention rates but rather represent inequities more accurately and, arguably, draw attention to racial or racist patterns of deprivation rather than family structures or culture. The possibility that families from ethnic minority groups may remain in disadvantaged neighbourhoods because of valuing cultural normality means it is vital that further research examines family socio-economic circumstances rather than using neighbourhood deprivation as a proxy. However, the very large inequities between groups, both in respect of which children grow up in highly disadvantaged neighbourhoods and which children end up on CPP or separated from their parents, will remain.

Deprivation and intervention rates: multiple ethnic categories

The broad ethnic categories hide as well as reveal inequalities in the application of child protection services. Tables 7 and 8 show rates of intervention for the eighteen ethnic categories in quintile 5 where numbers of children are ten or over (full tables in Supplementary material, Tables S5 and S6). Care should be taken in making comparisons outside quintile 5 because of low numbers.

The detailed data show important differences compared to the patterns for the broad ethnic categories. First, White British children in quintile 5 are shown to have had higher CPP rates than Mixed White/Caribbean or White/African children. It is the ‘Mixed White/Asian’ and ‘White/Other’ groups that had the highest rates. Second, within the Asian group, the relatively low rates of children categorised as Indian are of particular interest. Third, the ‘Black Caribbean’ category had a much higher rate than the ‘African’ and ‘Other Black’ categories.

Table 8 presents the equivalent data for LAC. Again, important distinctions are apparent. First, in quintile 5, the highest rates are seen in the ‘Other Mixed’ group and the ‘Mixed: White/Asian’ group. The rate for ‘Mixed White/Caribbean’ children in Q5 is similar to the rate for ‘White British’ children. Second, the LAC rate for children of an Indian background is low compared to the other Asian groups and much lower than other groups generally. The rate for White British children is almost nine times that for Indian children in Q5. Third, the high LAC rate for ‘Black Caribbean’ children in quintile 5 stands out by comparison with ‘Other Black’, ‘White’ and ‘Asian’ categories. Fourth, although the numbers are small, the rates for ‘White Traveller’, ‘Romany’ and ‘Other’ groups are notably high (Table S6).

When Tables 7 and 8 are compared, it can be seen that the patterns are not precisely the same. For example, more than twice as many Black Caribbean and ixed heritage children were being looked after as were on CPPs but only fifty per cent more White British children. For all the Asian groups, there was a higher proportion of children on CPPs than LAC. Again, this study cannot explain these very large inequities.

Our data were based on the numbers of children who were on a child protection plan or who were being looked after on 31 March 2012 and therefore cannot shed any light on how children moved through the system. Higher rates on a given date might result from higher rates of admission or longer stays on a CPP or being looked after, or a combination of factors.

Age, deprivation and intervention rates

Age is a further significant factor. Essentially, LAC rates increase with age while CPP rates decline. After the age of four, the combined CPP and LAC rate in the sample overall was fairly constant at between ninety and 100 children per 10,000. However, this also varied with ethnicity (Supplementary material, Tables S7 and S8). For CPPs, the gap between White and Mixed heritage rates, on the one hand, and Asian and Black children’s rates reduces with age, although, by the age of sixteen, numbers are very small. For LAC, the rates for minority ethnic categories showed a greater increase with age than the rates for White children. The effect is (Table 9) that the ‘gap’ between the proportion of White children subject to these state interventions compared to Asian and Black children narrows as age increases, when comparing equivalent neighbourhoods. The largest differences are amongst children under four. It has not been possible to analyse these inequalities for the smaller ethnic subgroups as the numbers are insufficient.

Discussion and conclusions

Ethnicity, like deprivation, has a powerful association with a child’s chances of experiencing a state intervention such as being placed on a child protection plan or being looked after away from their parents. If all children had the intervention rates of Asian children and similar deprivation patterns, there would have been 77 per cent fewer LAC in the West Midlands on 31 March 2012 and 58 per cent fewer children on CPPs. Equally, if all the ethnic minority children had had the same pattern of socio-economic circumstances as White British children, there would have been many fewer subject to intervention. The lack of attention paid to ethnic inequalities in public policy discussions of the child protection systems in the UK and Europe is striking because of the profound implications both for social justice and for public expenditure. However complex the issues, not to pay attention to these life-changing differences in state interventions in family life is unsupportable.

Of course, the fact that one ethnic category has a lower intervention rate than another is not necessarily evidence that the consequences for children are better (or worse). Internationally, child protection systems lack agreed measures to determine whether higher or lower rates imply better outcomes for children or lower long-term costs for society. The evidence is not available. It should also not be assumed that rates for all children could be reduced to the rates for the lowest sub-group without the potential for harm just as it cannot be assumed that rates should be raised to those of the highest on the grounds that children’s needs are being missed. The fact that we cannot answer these profound challenges is deeply problematic.

A number of important qualifications to this research must be recognised. First, this is the first time that such detailed evidence has been presented, breaking down broad ethnic categories into subgroups and controlling for neighbourhood deprivation. The work needs to be replicated, ideally with a larger sample. Second, the evidence would be much stronger if it was based directly on family socio-economic circumstances rather than using neighbourhood deprivation as a proxy. The relationship of family circumstances to neighbourhood deprivation may well be different for different ethnic groups. Third, Census and Office for National Statistics data (ONS, 2011) suggest that the validity and consistency of application of the ethnic categories in which official data is recorded and reported require attention if public policy is to be based on the findings. It would be very surprising if African children of Somali heritage had the same intervention rates as those from Nigeria or Uganda or Zimbabwe, although it seems clear that Black families with an African-Caribbean heritage have a very different relationship to state welfare services from those of direct African descent. Fourth, further work needs to be done to look at how children from different communities, at different ages and in different circumstances progress through the child protection system to test whether the inequalities reported here relate to who enters the system or how the system responds. Fifth, as adoption and special guardianship orders have increased radically in recent years as mechanisms by which children leave the LAC statistics, and kinship care is a vitally important additional form of substitute care, the examination of racially based differences needs to be expanded to include these forms of state and family intervention. Without doing so, the complete picture of how the lives of children are affected will not be available.

The significance of ethnicity as a factor and the different patterns for different ethnic categories might be taken as undermining the case that socio-economic structures are of primary importance in explaining inequalities in intervention rates but this is not a valid conclusion. It is rather the case that both racial and economic structures, and their interaction, are of central significance. Indeed, the data on ethnicity strengthen the socio-economic case, as each ethnic category shows the same positive correlation between increasing neighbourhood deprivation and increasing intervention rates, whenever numbers are sufficient. This applies in almost every case, not only for each of the eighteen ethnic categories, but within each age group for the five broad groups. To assess or inspect a local authority’s intervention rates without taking into account both the ethnic and socio-economic distribution of the child population would make little sense and would lead to inaccurate conclusions.

The data presented here demonstrate that understanding what is happening to children and families subject to powerful state interventions requires action on a number of fronts; official data, research, practice, training, inspection and policy making. We need better knowledge. More detailed and consistent information needs to be collected, reported, analysed and published than has hitherto been produced through official statistics. Four obvious developments would provide the basis for informed policy making. The first would be to collect or enable linkage with data on the socio-economic circumstances of families with whom the state intervenes. Such data are collected for the analysis of the health and education systems in the UK—why not for children’s services? Second, careful thought should be given to reviewing the ethnic categorisation system for official data on children’s services and determining which level of categorisation would be appropriate for different reporting purposes within the system. Detailed categories might be required for some purposes while broad categories may be useful for others. Third, the analysis of data should include mechanisms for reporting how children move through the system. The different ratios of CPP to LAC for different ethnic categories require data as a basis for understanding. Fourth, the prevalence rates for children from different ethnic groups should be produced and published.

It is only by understanding the subtle but powerful interaction of racism, ethnicity, socio-economic circumstances, service provision and wider social policies that it will be possible to make sense of inequities in intervention rates between ethnic groups and this is an important agenda for research. Research has a different role to that of official data, in particular to theorise and test explanatory models of intervention inequities. Qualitative methodologies are a necessary adjunct to quantitative evidence in order to capture the attitudes and actions of actors in the system, children, parents, social workers, workers from other professions and agencies, the legal system and policy makers.

Those involved in direct practice and decision making (and training) should also pay much closer attention to patterns of intervention, process and outcome. It is as a result of a myriad of individual actions that differential outcomes emerge, decisions affecting the patterns of service provision offered, who accesses services, how families are assessed, what assumptions underlie subsequent interventions or non-interventions, and the short- and long-term outcomes. There is a continuing risk of a colour-blind approach through which the underlying patterns and their consequences are allowed to become invisible.

One important theme of this article has been to present evidence which requires rethinking the long-held assumption in the UK that a higher proportion of Black children are looked after than White children. Once controlled for deprivation and for more specific ethnic categories, that assumption no longer holds in that simple form. Explanations for the disparities in intervention rates between White, Black and Asian children based on crude assumptions about family patterns or parenting need to be informed by the differential exposure to socio-economic disadvantage. It remains the case that—overall—Black children are more likely than White to experience separation from their parents through state action, because so much larger a proportion of Black children compared to White live in very disadvantaged neighbourhoods. Overall, Black children of Caribbean heritage were more than twice as likely as White British children to find themselves looked after, in our sample, but not more likely to be on CPP. Black children of Caribbean heritage were almost 40 per cent more likely than White British children to be looked after even in the most disadvantaged quintile of neighbourhoods where over four-fifths of them lived. However, rates for Black children of African heritage were much lower than those for either Black Caribbean or White British children, even after controlling for deprivation, and this too requires further detailed investigation.

Our evidence must not be interpreted as suggesting that, because controlling for deprivation has such as dramatic effect on the picture, this is no longer an issue of importance. Rather it has revealed even greater evidence of profound inequities structured by race and socio-economic structures. Equally, simplistic assumptions that ‘Asian’ (extended) families provide a better protective context for children to grow up in require examination against the Mixed experience of the sub-categories. There is a major role for research in developing and detailing understanding of such large differences in interventions and their outcomes.

The political and practical complexities involved in examining the very large and persistent inequalities in the chances of children from different ethnic categories receiving a child protection intervention should not be a barrier to action. In light of our evidence, much more data should be provided through the reporting of analysis of official data on intervention rates at the local authority level and below. Action should be taken to determine whether the inequalities in intervention rates by ethnic category should be a focus of policy, as they are in education and health. Inequities between ethnic categories are not a matter of a few percentage points of difference, but many multiples. This is not acceptable in the absence of an understanding of the reasons or of the overall outcomes for children. Moreover, there may be important lessons to be learnt from minority communities with low rates about how best to protect children outside state care. Differential patterns in out-of-home care and abuse and neglect are worldwide and long-standing. It is more than time for all those involved in or concerned about child welfare to ask serious questions about the practices, policies and wider social structures which maintain such unjust inequities.

Acknowledgements

We wish to thank the Nuffield Foundation (grant reference: CPF/41337) for their support in funding the project and the participating local authorities who provided the data and contributed to the development of the project through discussion and engagement. Without their positive desire to improve their services, this project could not have been completed. The findings, interpretations and conclusions expressed in this paper are entirely those of the authors.

References