CHILD POVERTY, EMPLOYMENT AND ETHNICITY IN THE UK: THE ROLE AND LIMITATIONS OF POLICY

In this paper, I discuss two major policy agendas of the current UK administration: one concerning child poverty and another concerning the employment of ethnic minority groups. In this introduction, I briefly set out the context for each and argue that, despite their prominence, there has been little direct connection made between the two. This is despite the much higher poverty risks for children living in minority group households, and is, I suggest, partly a result of the way the two agendas are conceived – and consequently investigated. The two agendas are both subject to substantial monitoring and targets; and a similar process is seen in EU level reporting structures, established following the 2000 Lisbon poverty goal. Yet though such requirements are comprehensive and pick up multiple related areas, they do not necessarily stimulate connections being made across them.

In Section 2, I present a descriptive analysis of child poverty and ethnicity; and I go on to consider how we might understand its causes (Section 3), focusing particularly on employment-related factors. After setting out some messages for policy from this analysis, the paper concludes by a review of where evidence on the connection between ethnic employment disadvantage and child poverty is insufficient. I argue that, given the wealth of investigation both into child poverty and into ethnicity and employment, this lack of research evidence is a consequence of the way that child poverty and ethnic minority employment disadvantage are constructed and measured.

In 1999, the British Government stated its resolve to end child poverty (Blair 1999). In doing so it recognised both the extent of child poverty in the UK and the moral imperative to do something about it (Walker 1999; Platt 2005). In his 1999 lecture, ‘Beveridge Revisited: a welfare state for the 21^st century’, the Prime Minister declared that

This aim was re-iterated by the Chancellor of the Exchequer, both in speeches and in a range of adjustments to the benefits system designed to favour families with children. The aims to reduce child poverty were structured into incremental targets for 2005 (reduce child poverty by a quarter) 2010 (halve child poverty) and 2020 (abolish child poverty) and supported by annual monitoring of a range of measures (DWP 2003, 2006a) to establish how far the ‘historic aim’ was being achieved.

While a national aim, the child poverty ambition is also connected to the development of a Europe-wide agenda for addressing child poverty (Ruxton and Bennett 2001) and to the Lisbon goals more generally. The European Council that took place in Lisbon in 2000 agreed that member states should make a decisive impact on poverty by 2010. Here, too, the temporal target was succeeded by the establishment of a national reporting structure; and common indicators were established in 2006 (European Commission 2006). Poverty risks, including those for children, form primary indicators; and there are clear overlaps between the national EU reports in structure, emphasis and the measures they use.

A distinctive and persistent feature of discussion and debate on child poverty has been its construction primarily as an end in itself, alongside the cross-national – and within the UK cross-party – acceptance of this as a valid and important end (Hirsch 2006). Thus, child poverty reduction is seen as a responsibility of the state – a ‘duty’ owed by the state to children living in it, and the terms of debate revolve around this assumption.

In UK strategies to reduce child poverty, lone parent families are a particular focus, as are those in workless households. The additional risks associated with large families and minority ethnicity are also mentioned, but only briefly and without a comprehensive strategy (DWP 2006a). The UK's EU monitoring report on social exclusion and social protection also emphasises the child poverty risks associated with lone parenthood, but no mention is made of ethnicity in relation to the substantial discussion of Policy Objective 1 on reducing child poverty (UK Government 2006). Given that part of the Lisbon agreement was concerned with cooperation and sharing good practice, this segmentation indicates a lack of common concern with integrating understanding of discrimination and poverty indicators at the EU level or among individual states.

At around the same time that the child poverty agenda was established, the British Government's Policy and Innovation Unit began work on an examination of the position of ethnic minorities in the labour market. This initiative was informed by an accumulation of evidence on the large differences in employment rates between ethnic groups. It was stimulated both by the dominant ‘welfare to work’ agenda and by a heightened sensitivity to race relations and structural inequalities, following the publication of the Macpherson report (1999), which established that institutional racism was a feature of the UK public sector. Following the publication of two reports (PIU 2001, Strategy Unit 2003), a cross-departmental ethnic minority employment task (EMETF) force was established in 2003, alongside a Public Services Agreement (that is, funding related) target on reducing the ethnic minority employment gap. A number of initiatives have followed in relation to: outreach for minority ethnic groups, procurement, engaging unemployed partners of claimants and using area initiatives to focus on minority groups (DWP 2004, 2006a; UK Government 2006).

Yet, despite some stress on workless households (Ritchie et al.2005) and the recognition of the propensity of partners of unemployed to themselves be without work (Gregg and Wadsworth 1996), the majority of the focus has been about individual adults gaining, or being supported to gain employment. If the emphasis in the child poverty agenda is on the duty the state owes to the child, then in the employment agenda, the emphasis has clearly been, throughout the ‘welfare to work’ drive, on the responsibility of the individual to find and take up work (DSS 1998). At the level of EU reporting structures, ethnic minority disadvantage is also closely linked to employment, but is predominantly considered under discussion of discrimination, largely separating the experience of minorities even from the mainstream employment agenda and reflecting instead concerns with ‘community cohesion’ (UK Government 2006).

Research evidence on employment is carried out at the level of the working age individual, which will thus tend to imply individual-level solutions and responses. By contrast, poverty, and thus child poverty, is measured at the level of the household, which implies looking at the overall household composition. Thus, there is a lack of coincidence between the ways the problem and solutions are framed. Additionally, though the basic differences in poverty between ethnic groups are well known – many of the instruments that are used for monitoring are not (and are often not able to be) sensitive to ethnic group differences.

The lack of integration between the ethnicity and child poverty agendas is just beginning to be recognised – at least at the national level. In 2005, a parliamentary committee investigation into child poverty recommended greater attention to ethnicity; and, more emphatically, in 2006 the EMETF invited a paper on ethnicity and child poverty (Platt 2006a). This paper was explicitly intended to provide a background to the work of the EMETF and thereby incorporate child poverty into the ethnicity employment agenda. Material from this paper was simultaneously included in the Department for Work and Pensions’ specially commissioned review of their child poverty strategy (Harker 2006).

In this article, I draw heavily on that paper, but develop its argument in relation to the possibilities of and conditions for an integrated understanding of poverty and ethnic minority employment.

In illustrating the extent of poverty among children according to ethnic group this paper employs an income measure of poverty. While the use of income as a measure of poverty is not uncontested (Ringen 1988; Pantazis et al.2006) it has the advantage of transparency – it is clear what is being measured; and the particular measure of 60 per cent of median income as the low income threshold has been accepted throughout Europe as a core monitoring measure. Moreover, income measures may be more appropriate for ethnic group comparisons as they do not make normative assumptions about what the income is actually spent on (Platt 2006b).

In looking at breakdowns of poverty, we can distinguish between differences in shares and differences in risks of poverty for children living in families with different characteristics. For example, in the population as a whole lone parent families account for around a 40 per cent share of poor children, that is a minority of children; but children in such families face much higher risks: around 50 per cent are poor, compared to 20 per cent overall.

Policy tends to focus on the situations of higher risk as, first, being more likely to reach poor children: if you target lone parent families, one in two will be poor, whereas if you target all families only one in five children will actually be poor. Second, the higher risks are seen at some level as being more ‘unfair’. Why should particular family characteristics dramatically increase a child's risks of growing up in poverty? The same arguments can be applied to children's chances of poverty according to the ethnic group of their family, where minority group children make up 12 per cent of the population of children but, because of the much higher poverty risks they face, they make up 20 per cent of poor children.

However, as we see from Table 1, below, those aggregate minority figures mask great differences between children in individual groups. Recent figures from the Family Resources Survey (Table 1, columns 1 and 2) reveal that rates of poverty among Black African, Pakistani and Bangladeshi children are more than double the rate among white children, reaching over 72 per cent of Bangladeshi children, when measured after housing costs (or 60 per cent when measured before housing costs).¹ Rates of child poverty are also up to 10 percentage points higher for those living in Black Caribbean and Indian households. Taking child poverty seriously, then, should involve recognising the very different risks of being in poverty according to ethnic group.

If we compare the rates of poverty among households with children (Table 1, columns 3 and 4), we see a similar pattern; but the differences in child poverty rates are more extreme than the rates among households with children, given that the households with the highest child poverty rates also tend to have higher than average numbers of children. The point is an obvious one but relevant for policy in that employment policies tend to focus on the family unit, whereas child poverty calculations consider the number of children.

The next illustrations examine the variations in child poverty by ethnic group for those living in different sorts of household: lone parent and couple parent families; households with and without earners, households with and without disabled adults; households with varying numbers of children. Returning to the discussion of risks and shares, we can explore differences in shares of groups that are made up of particular family types, and the differences in risks between groups for a given family type. We can investigate the extent to which, for example, lone parent families make up a greater or lesser extent of poor children for any given group, thus leading to considerations of the extent to which policies targeted on particular family types are likely to reach that group. And we can also inspect the extent to which the risks associated with living in a lone parent (or other type of) family differ between groups. We might expect the risks associated with a particular family type to be largely independent of the prevalence of that family type within the group. That is, we would not necessarily expect the risks associated with living in a lone parent family to be either greater or smaller where lone parent families were very common (or vary rare). Indeed, in the absence of an ‘ethnic penalty’ (Heath and McMahon 1997) – that is, disadvantage associated with an ethnic group that cannot be attributed to relevant characteristics – we might expect the risks associated with particular circumstances to be broadly similar across groups. However, the prevalence of particular family types (in combination with the risks) will affect the share of children in poverty from that household type. For example, if 90 per cent of children live in couple parent families then we would expect the majority of poor children also to live in such families, even if the risks of poverty from living in a lone parent family were much higher.

Tables 2 and 3 are, therefore, informative about the potential effectiveness of interventions targeted at particular family types for different ethnic groups. Turning first to Table 2: among white children, those living in lone parent families make up the largest share of those in poverty, but the risks are highest among those living with couple parents but with no full-time worker.² Conversely the risks associated with living in a family in which there is at least one earner are low; but because such households make up the majority of households containing white children, they still account for nearly half of all white children in poverty.

By contrast, lone parent families account for only a small share of Indian, Pakistani and Bangladeshi children in poverty, but they make up over two-thirds of Black Caribbean and Black African children in poverty. The risks for all the ethnic minority groups are high for children in these circumstances (between 55 and 63 per cent chance of being in poverty); but they are not as high for the risks for those living with couple parents where there is no full-time earner, which top 80 per cent across the ethnic minority groups. Distinctively, though, for Pakistani and Bangladeshi children the risks of living in a couple parent family with at least one full-time earner are also relatively high (at 46 per cent). This is drawn out further in the final column, where we see that households with one or more earners carry a risk of poverty of over 50 per cent for Pakistani and Bangladeshi children. The risks for Indian children in such households are much lower (at 24 per cent) but they account for nearly two-thirds of poor Indian children.

Similarly, both risks and distributions vary when we consider households with different numbers of children and households containing a disabled adult (Table 3). Thus, we see that the majority of poor white, Indian and Black Caribbean and Black African children live in one- or two-child families, but over two-thirds of poor Pakistani and Bangladeshi children live in families with three or more children. For all ethnic groups, the risks associated with a larger family are higher than for those in a smaller family; but for Pakistani, Bangladeshi, Black Caribbean and Black African children the risks of poverty in a smaller family are still higher than those for white children living in a large family. When we consider the disability status of the household, we can see that living in a household with a disabled adult increases the risks of poverty across groups (compared with the overall child poverty rates in Table 1), except for the Black Caribbeans and Black Africans, where the share of child poverty made up from such families is also very small. However, there is still substantial variation between groups. It is not simply the fact of having more families with a disabled adult that increases the poverty rates for Bangladeshi children; their risks of poverty are over twice those for white children, at 83 per cent compared to 36 per cent.

In order to reduce child poverty among ethnic minority groups, and differentials between ethnic groups, it would be possible to focus on either those situations in which risks of child poverty are disproportionate (for example, Bangladeshi children living in households with a disabled adult), or situations that account for the greatest proportion of child poverty for particular ethnic groups (for example, Black Caribbean and Black African children living in lone parent families) – or those situations where they overlap (for example, Pakistani and Bangladeshi children living in households with at least one earner). In fact an approach which addresses both those households with higher risks and those where child poverty concentrates is likely to be necessary to reduce child poverty among ethnic minority groups. As noted, higher risks can be thought of as a form of ethnic penalty at the level of the household or family type, while distributions in family types that command a high risk can be associated with demographic, cultural and structural processes that result in these particular distributions. Thus, focusing on disproportionate risks will tend to imply policies targeted toward particular ethnic groups and the specific circumstances which lead to those increased risks. On the other hand, targeting risky situations which are disproportionately experienced by particular ethnic groups (such as workless households) may imply either (a) universal policies which attempt to address these particular situations (either the fact of workless households or the fact that they are likely to bring poverty), or (b) targeted policies if those situations (e.g., lack of work) are related to obstacles specifically faced by particular groups.

Section 4 returns to the question of policy implications. The next section examines some of the factors that contribute to these differential risks.

As noted, it is household income that is relevant to whether or not children living in that household are judged to be in poverty. But employment rates and ethnic penalties in employment and pay are typically investigated at the level of the individual (e.g., Blackaby et al.2005) or occasionally the couple (e.g., Dustmann and Fabbri 2005); and separate analyses are not typically carried out for those with and without children.

Nevertheless, information on the employment experience of different ethnic groups will go some way towards helping us understand ethnic differences in child poverty. At the individual level there are clear differences in unemployment and economic inactivity across groups, with men and women from all the main ethnic minority groups having higher rates of unemployment than majority men and women (Figure 1). In addition, minority group men and women have higher inactivity rates than white British men and women (with the exception of Caribbean women compared to white British women, where the rates are the same).

For those in work, pay also differs with ethnicity. Figure 2 shows hourly pay for those in full-time work by ethnic group, for men and women. The lower rates of pay for the ethnic minority groups, with the exceptions of Indian men compared to white British men and of Caribbean and Indian women compared to white women, is evident. The particularly low pay of Bangladeshi men stands out; and this is without taking account of the pay deficits associated with part-time working (Platt 2006c). As Figure 1 illustrates, rates of part-time work also vary by ethnic group, with Bangladeshi men having the highest rates among men. These figures on hourly pay also do not include self-employment, which varies greatly by ethnic group: Pakistani men have particularly high rates and Black African men having lower than average rates (Figure 3).³ Income from self-employment is hard to measure reliably, but much self-employment among ethnic minority groups can be regarded as representing a constraint rather than a choice (Clark and Drinkwater 1998, 2000), and can bring poor conditions, insecurity and relatively poor returns.

While differences in characteristics other than ethnicity contribute to the employment disadvantage of certain minority groups, abundant research reveals an ethnic penalty in employment for most minority groups (for example, Heath and McMahon 1997; Berthoud 2000; Blackaby et al.2005; Heath and Cheung 2006). This penalty affects even those groups, such as Indian, that appear to be performing quite well in the labour market: the finding of an ethnic penalty indicates that they should be performing even better. For the other groups, the ethnic penalty accounts for part of the observed employment disadvantage. This suggests that concentrating solely on the characteristics that improve the employability of ethnic minority group members, while it may be part of a solution, will not eliminate the employment gap.

The employment gap experienced by minority groups and the ethnic penalty (including discrimination), which contributes to it, therefore contribute to child poverty. The relationship is not, however, straightforward. The impact of employment disadvantage on child poverty depends on how earnings from employment aggregate at the household level and how they are, or are not, supplemented by other sources of income. There is, therefore, not a direct relationship between employment disadvantage and child poverty. For example, the Indian group performs relatively well in terms of employment and pay, but rates of child poverty are substantially higher than for white children (see Table 1).

To understand child poverty rates, we therefore have to consider how earnings and other sources of income cluster at the household level in households containing dependent children. And we then need to consider the demands that are placed on that income and explore how these differ between households according to ethnic group.

Section 2 showed how different types of household contributed different shares of the total child poverty for any given group. That implied that differences in rates are, in part, driven by the prevalence of different sorts of households across groups. Household characteristics which are relevant to child poverty and which have been shown to vary by ethnic group are: household composition and demographics (Dobbs et al.2006); complex (multi-family) households (Owen 1996); large families, with large numbers of (dependent) children (Bradshaw et al.2006; Iacavou and Berthoud 2006); and lone parent families (Berthoud 2005; Lyon et al. 2006).

To illustrate the impact of these differences in family structure, we can examine dependency ratios, that is, how many dependants (both children and older people) there are per working age adult. Variation simply in these dependency ratios by ethnic group in households with children will cause child poverty rates to vary across groups. Table 4, column 1 shows the availability of working age adults per dependant (older or younger) in working age households. There are, on average 1.4 working age adults per dependant, but this varies by ethnicity from 1.2 among Black African households to 1.6 in Indian households.

Additionally we can consider the proportion of working age adults in any household who are actually employed. (This relates to Table 2.) Column 2 of Table 4 provides the ratio of adults in work to both children and older adults in households containing children under 16. While there is one adult in employment for every dependant in white and Indian families with children, there are, on average, only 0.5 workers per dependant in Bangladeshi families and less than one worker per dependant in other ethnic minority group families. This, however, excludes any non-working adults of working age that those adults in work may need to support. Thus, the ratio for Indians of one working adult to one dependant, which appears comparable to the white group, does not take account of the fact that average numbers of working age adults are higher in Indian than in white households (column 1). Columns 3 and 4 make a similar point by showing the proportion of households with dependent children that have no one in employment. Black African households with children are particularly likely to lack someone in employment, partly as a consequence of the large proportion of lone parent Black African families.

Living with a sick or disabled adult can also affect both household employment levels and child poverty. We saw, in Table 3, that the proportions of poor children who were living in a household with a disabled adult varied substantially by ethnic group. The final columns of Table 4 show the variation in proportions of households with children that contain an adult who is long-term sick. Such households have higher risks of poverty (Martin and White 1987; Burchardt 2000), which will tend to lead to higher rates of child poverty where they form relatively high proportions – i.e., among Pakistani and Bangladeshi families. The presence of someone who is long-term sick is likely to impact on the labour supply of other household members – they may need to be available for care; and where the person with chronic illness is of working age, the illness is also likely to impact on their own labour supply (Berthoud 2006).

For those households where there are adult(s) in work it is also important to consider that the returns from that work in terms of pay may not be sufficient to lift the household out of poverty, given the pay differences illustrated in Figure 2. For those households with children without (all) working age adults in employment, or where rates of pay are low, other sources of income will be critical in preventing child poverty or ameliorating the depth of poverty. In multi-generation families the extent to which the older members receive pensions and other benefits will be relevant (Burton 1997; Ginn and Arber 2001; Gordon et al.2002). (Although one of the reasons for the co-residence may be to protect the older adults from poverty in the absence of pension entitlement.) Take-up of state benefits, including disability benefits, by other household members is also important. Evidence indicates that there may be substantial differences in benefit take-up between ethnic groups: investigations into rates of receipt (e.g., Salway et al.2007) and preliminary investigations of take-up rates by ethnic group have suggested systematic differences. There are also indications that ethnic minority groups experience both more limited entitlement to certain benefits (through for example interrupted contributions records) and evidence from qualitative studies has shown that minority groups may be less likely to claim various forms of benefit to which they are entitled (Law 1994; Bloch 1997; Gordon et al.2002; Barnard and Pettigrew 2003; Platt 2003).

Access to savings and assets also provide a cushion against poverty. Once again the evidence is not extensive on ethnic group differences in savings and assets among families with children, but what there is suggests that ethnic minority group families have fewer of such resources to draw upon (DWP 2006b).

Finally, the extent to which children themselves remain actually or effectively dependants following the end of compulsory schooling will also affect child poverty rates – both their own and those of younger siblings. We know that children from ethnic minority groups tend to remain in education for longer than average (Drew et al.1997; Bradley and Taylor 2004; DfES 2004). Thus, though of ‘working age’ they may not be available to work, and will require support for a longer period from the household as a whole. This may be a positive development in the long term – education is reasonably effective in ensuring better outcomes – but it may be bad for poverty risks of these dependent or ‘semi-dependent’ children in the immediate term and also have consequences for their younger, dependent siblings.

Linking the information on child poverty and ethnicity is important if effective policy to mitigate the extreme poverty risks experienced by children from some minority groups is to be developed. That policy will be informed by analysis and research to the extent that it is available; and I consider gaps in research briefly in Section 5. Here, I briefly outline some of the clear policy messages arising from the preceding discussion.

In general, employment policy has an important role to play to respond to the challenge of extremely high child poverty rates among certain minority groups. It is clear that lack of employment among working age adults is a major contributor to child poverty among ethnic minority groups. The nature and construction of income maintenance policies are, however, also a factor in contributing to or undermining child poverty.

As mentioned in Section 2, ethnic minority poverty can be tackled in part through general policies which have a particular bearing for certain groups. For example, lone parent policies will disproportionately affect Black Caribbean and Black African families, while policies directed at reducing the poverty of large families will disproportionately affect Bangladeshi and Pakistani families. Targeted policies can also play a role. Policies targeting racial discrimination in employment are an obvious example. Both universal and targeted policies are implicated by the preceding analysis. Specific examples of universal policies might include those supporting people with health problems to remain in work (e.g., DWP 2006c); affordable childcare policies (e.g., HMT et al.2004a) and the development of skills among the unemployed (e.g., HMT et al. 2004b). The financial support offered to those who remain in post-compulsory education for long periods, acquiring or upgrading qualifications also merits further attention in the light of the longer staying on rates among minority ethnic groups. Educational Maintenance Allowances alleviate some of the financial burdens for those aged 16–18, but not beyond this age.

In addition, ensuring a reasonable level for the minimum wage will disproportionately affect low-paid workers. Low rates of pay in part-time work would appear to contribute to family poverty. Equalising terms and conditions of part-time compared to full-time pay does not solve the part-time pay deficit, given the very different jobs that are available part-time compared to full-time (Manning and Petrongolo 2005). Instead flexibility within employment which allows reduction of hours when circumstances (such as caring responsibilities or health status) require it are important for maintaining reasonable levels of pay alongside part-time hours.

Targeted policies should include ensuring the cultural accessibility of child care to those from different ethnic groups; and promoting the effective utilisation of skills qualifications and experience among job seekers who migrated to the UK as adults. Given the evidence on limited career progression faced by minority ethnic group workers, there is a need for governments to find ways of engaging effectively with within-employment discrimination as well as job entry discrimination.

Such policies are all likely to play a part in increasing the amount of earnings coming into minority group households with dependent children. But earnings from employment are only one element of household income for households with dependent children, albeit the most important. State benefits can also increase the financial resources available to families with children. This extends to the benefits that are or are not received by all members of the household, not just those directly related to the presence of a child (such as child tax credit). Despite the lack of systematic studies of ethnicity and take-up, evidence points to both differential take-up among those apparently eligible and, in some cases, exclusionary eligibility criteria. Thus, attention should be paid to ensuring take-up among the eligible in families with children and to scrutinising the impact of benefit rules on particular groups.

Pensions policy and benefits for older people will affect multi-generational households with children, where the family members use co-residence to pool risks. The separation of different life stages cannot be assumed, even though this is another example where the European Lisbon framework makes a clear distinction. Such multi-generational households are more likely to be from minority groups, and therefore accumulated disadvantage across generations can cluster in such families with children. Sickness related benefits are potentially an important component of income for those households with children containing members with long-term illness. In Tables 3 and 4 we saw striking differences between the proportions of such families by ethnic group.

In addition, a higher level of financial remuneration for those who take on caring responsibilities for the long-term sick could have positive effects on child poverty rates and facilitate the combination of such caring with child care, arguably promoting better welfare consequences for the family in certain circumstances than attempting to place the ‘carer’ in paid employment.

Despite this focus on employment and income, it is also important to recognise that there may well be trade-offs between household income, especially income gained through employment, and child welfare more broadly conceived. For example, where there is a chronically sick adult in the household, the non-employment of another adult may serve to ‘protect’ children from some caring responsibilities or more effectively manage the household stress that such a situation can give rise to. In addition, parents are not necessarily interchangeable, and the loss of employment of one parent will not necessarily ‘free up’ the other parent to engage in paid work. Having a main carer parent at home (in both lone and couple parent families) may provide more stability, albeit on a low income, than a having a parent in poorly paid work, particularly if there are risks of ‘cycling’ in and out of insecure job opportunities (Evans et al.2004; Platt 2006d). There may not be a single view among parents as to how to maximise the welfare and the opportunities of the next generation, particularly among those ethnic minority group members facing substantial current constraints in relation to available jobs and pay (Dale et al.2002). There needs to be more understanding and recognition of the potential impacts on children of unstable patterns of work and care.

That much of the discussion above has been tentative is based in part on lack of sources for detailed information about the causes of differential poverty rates between ethnic groups (an information ‘gap’), and in part on a mismatch between a very rich seam of research into ethnicity, diversity and disadvantage and our understanding of child poverty (a gap in the analytical approaches used). In this final section, I consider both the information gaps, where we lack data to understand or adequately to explain ethnic group differences and analytic gaps, where the data exist but approaches which enable us to engage with child poverty and ethnicity have not been fully developed.

Policy-relevant research does not exist in a vacuum, but is shaped by policy discourses and approaches. This can happen both directly, through commissioned research, but also indirectly where concepts that pervade policy discussions are taken up and explored, even if they are challenged in the process. We can think for example of the surge in UK investigations into social exclusion that followed the Labour administration embracing it as policy priority in 1997. Of course, the relationship also works the other way, with policy absorbing and adapting research and its approaches. The symbiotic relationship can be fruitful, but it also risks narrowing the research focus into what can be considered acceptable territory. This may be a partial explanation for the wealth of labour market analysis by ethnic group, both commissioned and produced for purely academic purposes, and the relative paucity of systematic investigations into poverty and ethnicity. For, as discussed, labour market evaluations place the individual at the heart of the analysis, their characteristics and constraints, and imply individualised interventions. By contrast analysis of household welfare is not susceptible in the same way to such individualised response, and requires a perception of the issue as being about inter-relations of circumstances rather than characteristics attached to individuals. It also requires, as I have suggested, a recognition of responsibility from the state, rather than stressing the responsibility of the individual, which has become an increasing focus of political discourse in relation to ethnicity (Blair 2006).

Recognising the information gaps and expanding analytic gaps can help to reshape policy discourses to consider policy concerns in new or more integrated ways. Thus, the proposals made here are not simply concerned with finding out more, important though that is, but also with using additional research to rethink the agendas.

One area that could be much more fully developed within current data constraints is a detailed examination of poverty across ethnic groups, in particular for families with children. Controlling for different types of household and household characteristics, analysis could examine not simply comparisons of proportions, as has been done in this paper, but also poverty gaps, variation according to family and household type, and the composition of income in terms of the contribution of different sources (earnings, benefits, etc.) and which members they originate from. Such information would enable much more focused understanding of the potential role of employment policies in impacting on child poverty (and ethnic variation) and the different weight that should be attributed to the various messages for policy already identified. Surprisingly little work has been done on these basic but important – and interesting – questions since Berthoud's (1997, 1998) early analysis.

More work could also be done on self-employment and its role across ethnic groups in contributing to household welfare and mitigating poverty. Income and earnings analyses typically ignore those in self-employment, given the difficulties of accurately estimating income. However, the differences in rates of self-employment by ethnic group render this an important area for further investigation at the household level to understand its impact on children. The long hours associated with self-employment and its frequent involvement of other family members may also be important in considering its role in the family, its impact on child welfare and how it constrains the labour supply of other family members.

It might also be possible to estimate ethnic employment penalties at the household level, for households with children. Building on analysis of individual ethnic employment penalties, this could investigate the interaction of characteristics within households and how they might lead to ‘unexplained’ variation between ethnic groups.

These areas would enhance the possibility for policy to recognise ethnic variations in child poverty as a part of a linked employment and child poverty policy agenda and to form more tailored responses. Nevertheless, there are a number of areas where investigation (and thus policy information) is not possible within current data constraints. The necessity for using pooled years of data for analysing ethnic group variation in itself makes regular monitoring or comparison over time problematic. Moreover, there is a substantial deficit in our ability to understand the role of temporal factors in poverty and employment difference. Long-term poverty is acknowledged to be potentially more serious than short-term poverty in terms of its effects on children – including effects well into the future (Jenkins and Rigg 2001; Ermisch et al.2001). Low-income dynamics are analysed by the Department for Work and Pensions (DWP 2005) are incorporated into national and European reporting frameworks. If poverty analysis by ethnic group becomes central to considerations of child poverty, then an obvious complement would be to understand how it operates over time. Moreover, longitudinal analysis enables some disentangling of causal relationships, such as whether families are or are not poor prior to changes in employment status of household members, relationships that may well show variation by ethnic group.

Currently, data restrictions inhibit analysis in this area. However, the development of a large new panel study over the next few years (see http://www.esrcsocietytoday.ac.uk) will substantially enhance the possibilities for analysis of ethnically differentiated poverty dynamics. Similarly, there are currently few possibilities for research on employment dynamics and durations. The expansion of data linkage between administrative and survey sources, where the detailed ethnic information offered by the survey can be complemented by the detailed duration data offered by ongoing administrative records may soon enable analysis of moves into and out of registered unemployment. Such data developments are currently ongoing and if realised would greatly expand our very limited understanding of these dynamics by ethnic group. Analysis of poverty and employment dynamics by ethnic group would also be informative about levels of insecurity or instability, and thus the different experiences of poverty among children from different ethnic groups.

Finally, there are two areas where investigation could potentially resolve the extent or scale of factors which are suggested by existing research and widely accepted as being concerns for both income and employment experience, but where it is not yet possible to put a number on them. These are ethnic group differences in benefit take-up and the extent of employer discrimination. Lacking such a number is problematic both for evaluating the amount of weight to be placed on these areas and to disentangling the processes by which they occur. There are clear indications that the Department for Work and Pensions is now moving forward with the attempt to estimate non-take-up of means-tested benefits by ethnic group, using the methodology it currently employs to calculate non-ethnically-differentiated take-up rates. Clear evidence to evaluate the scale of employer discrimination, however, requires not existing survey data, but employer ‘tests’, such as those that were carried out in the 1970s and 1980s, with comparable ‘candidates’, varying only in their ethnic background, applying for jobs. Such ‘tests’ clearly raise employer sensibilities, and thus to be pursued they need commitment from research funders and conviction of their importance.

Ethnic difference should be taken into account in the current child poverty agenda. However, there are major challenges, both in conceptualising ethnic employment and child poverty and in taking forward an integrated agenda. The EMETF has made child poverty an underlying principle of its work: this gives scope for central attention to the greater levels of poverty that face children from certain minority ethnic groups. To do this effectively, however, far more attention must be paid to the household context in our conceptual and analytical approaches to ethnic disadvantage. This would help to reframe the policy agenda, so as to acknowledge the extreme risks of poverty faced by some children, how they arise and what interventions would be effective.

This paper is developed from a paper I wrote for the Ethnic Minority Employment Task Force. I has thus benefited from the contributions of a number of individuals in the Department for Work and Pensions. Many thanks to Will Driskell for proposing the paper and for offering comments, suggestions and the use of tailored HBAI figures; to Frances Goodwin for the HBAI analyses that form the basis of Tables 1–3; and to Carly Gray for allowing me sight of provisional work on benefit take-up and its variation by ethnic group. I am grateful to ONS for use of the Labour Force Survey and to the Data Archive at the University of Essex for making it available. Neither ONS, nor the Data Archive bear any responsibility for the analysis or interpretation offered here.

Lucinda Platt is lecturer in Sociology at the University of Essex. Her research focuses on child poverty and on ethnic minority disadvantage in the UK.