DEFINING AND MEASURING SOCIAL EXCLUSION: A CRITICAL OVERVIEW OF CURRENT PROPOSALS
The term 'social exclusion' played almost no part in Labour's pre-election lexicon. Within months, in August 1997, it was a central concept. In December 1997, the Social Exclusion Unit (SEU) was set up, for two years in the first instance, based in the Cabinet Office and reporting to the Prime Minister. The aim of the Unit is to develop co-ordinated policies to address social exclusion, described as 'joined-up policies for joined-up problems'. It has no spending budget, since its purpose is to make recommendations to the contributory government Departments, with a view to directing existing funding more effectively. Part of the brief of the Unit for its 1998 activities was the development of key indicators of social exclusion, which could be used in evaluating Government policy - and presumably the success of the Unit itself. Work did indeed start on this, but was abandoned in the summer of 1998, with little immediate likelihood of it being resumed. The possibility of a wider audit of poverty and social exclusion has stalled the Unit's work on this topic. They now think it likely that the development of indicators will be based outside the SEU, possibly at the Office for National Statistics (ONS). There have, however, already been a number of government reports, including some from the SEU itself, setting targets which appear to be connected with social exclusion, as well as think-tank reports, notably from the Institute of Public Policy Research (IPPR) and the New Policy Institute (NPI), which specifically seek to develop indicators of social exclusion.
There are always two major problems with quantifying social phenomena, definition and measurement. In an ideal world, definition precedes decisions about measurement, or where direct measurement is impossible, choice of proxy indicators. Part of the difficulty of finding indicators of social exclusion is that there is no agreed definition either of the phenomenon itself or of its main causes. In the remit of the SEU, social exclusion is described as 'a shorthand label for what can happen when individuals or areas suffer from a combination of linked problems such as unemployment, poor skills, low incomes, poor housing, high crime environments, bad health and family breakdown' (SEU 1997). Other possible definitions are the 'inability to participate effectively in economic, social, political and cultural life, alienation and distance from the mainstream society' (Duffy 1995) or 'the dynamic process of being shut out ... from any of the social, economic, political and cultural systems which determine the social integration of a person in society' (Walker and Walker 1997:8). Social exclusion is in all these versions presented as a multi-faceted problem. It is related to poverty, especially understandings of poverty which go beyond low income and address the multiple dimensions of deprivation. But the nature of the relationship is not clear. Poverty and social exclusion may be analytically separated, with poverty being the 'lack of material resources, especially income, necessary to participate in British society' (Walker and Walker: 8). But some definitions of poverty include social exclusion understood as lack of participation in social life. Overall poverty, as defined by the 1995 Copenhagen World Summit on Social Development, involves:
lack of income and productive resources to ensure sustainable livelihoods; hunger and malnutrition; ill health; limited or lack of access to education and other basic services; increased morbidity and mortality from illness; homelessness and inadequate housing; unsafe environments and social discrimination and exclusion. It is also characterised by lack of participation in decision-making and in civil, social and cultural life. (United Nations 1995:57)
Two related problems flow from these definitions. The first is the difficulty of separating out, even analytically, social exclusion and multiple deprivation. The second is that since both are necessarily multi-faceted, they require sets of indicators, rather than single ones. Which indicators are chosen, and which are seen as the most important, depends on views of both the nature of social exclusion and its causal links to poverty, which frequently remain implicit rather than explicit. To tease out the various meanings of social exclusion which are embedded in current political debate, Levitas (1998) develops a model which identifies three different approaches. Importantly, this is a model, and although there are tensions and contradictions between the approaches, they often coexist, however uncomfortably, in individual documents. The model can be used to explore these contradictions, as Watt and Jacobs (1999) have done with some of the output of the SEU. The contrasting approaches have very different views of the causal relations between poverty and exclusion, and different implications for key indicators.
The first of these approaches is a redistributive discourse (RED) which derives from critical social policy, and which sees social exclusion as a consequence of poverty. Thus Peter Townsend (1979, p.32) argued that poverty should not be understood in terms of subsistence, but in terms of people's ability to participate in the customary life of society: 'Individuals, families and groups can be said to be in poverty when ... their resources are so seriously below those commanded by the average individual or family that they are, in effect, excluded from ordinary living patterns, customs and activities'. Although 'resources' here does not simply refer to cash incomes - and importantly includes access to collectively-provided services - a central element in the RED approach has been that since social exclusion results from poverty, raising benefit levels to reduce poverty is crucial to reducing exclusion.
Much of current policy, however, is implicitly or explicitly rooted in a different model of exclusion, in which the key element, is labour-force attachment. This is underpinned by a discourse about social integration (SID) in which paid work is represented as the primary or sole legitimate means of integrating individuals of working age into society. The excluded are those who are 'workless', or, in the case of young people, at risk of becoming become so. Unlike RED, SID leaves little room for the reward of unpaid work through the benefit system, and glosses over the ways in which paid work may fail to prevent exclusion (for example by being poorly paid), or even cause it where long and asocial hours block other forms of social participation. While the lead indicator of social exclusion for RED is low income, for SID it is unemployment or 'economic inactivity' - a concept which intrinsically denies the value of unpaid, non-market work.
The third approach is a moral underclass discourse (MUD), which emphasises moral and cultural causes of poverty and which is centrally concerned with the moral hazard of 'dependency', and thus with workless households rather than individual labour market attachment. MUD tends to replay recurrent themes about 'dangerous classes' (Morris 1994, Murray 1990, 1999), to focus on consequences of social exclusion for social order, and on particular groups, such as unemployed and potentially criminal young men, and lone parents, especially young never-married mothers.
It is not easy to establish the relative weights of these different versions of social exclusion in the current debates over indicators. The call for a national audit of 'poverty and social exclusion' makes no separation between the terms. Nor does the Poverty and Social Exclusion (National Strategy) Bill introduced in Parliament under the 10-Minute Rule on 10 February 1999. The recent renewed emphasis on the link between 'poverty and social exclusion' might be seen to imply RED, but this is not necessarily so. MUD also posits a strong connection between poverty and social exclusion, but sees the causes of poverty as lying in cultural and moral (self) exclusion rather than the other way round. Problems of definition run through current attempts to produce appropriate indicators. This primary difficulty is linked to two others. Firstly, as mentioned earlier, both poverty and social exclusion require multiple indicators. This means it is possible to draw up a provisional set of indicators without clarifying the underlying definitions and causal relationships, and without any statement of priorities between indicators. Secondly, pragmatic considerations have led to an apparent consensus that the indicators used in a 'poverty and social exclusion audit' should be drawn from statistics, principally official statistics, which are already routinely collected. There are some persuasive arguments for this. Such an audit is more likely to be agreed if it requires little, if any, additional government expenditure. If it is to be in place quickly - particularly in time for the next election - and if it is to afford a comparison with the state of affairs inherited by Labour, it needs to rely on existing data sources. But this also means that rather than moving, as research ideally should, from definition to operationalisation to data collection, the process is largely reversed: we move from available data to an implicit definition embedded in the flawed data sets which already exist, and which never needs to be closely scrutinised. Of course, not all the discussion of potential indicators is theoretically naive, but it is very strongly constrained by the pragmatics of immediate influence on policy. This is true of both the major think-tank reports discussed below.
Monitoring poverty and social exclusion
The best advice the SEU could offer in response to an enquiry about potential indicators was to direct me to a recent report from the New Policy Institute (NPI), Monitoring Poverty and Social Exclusion: Labour's Inheritance (Howarth et al. 1998). The NPI report (funded by the Joseph Rowntree Foundation) is an impressive-looking document, with a lot of glossy graphs and charts. It is intended to form the basis for a regular official report on poverty and social exclusion, whether this is eventually produced by the ONS or an independent agency. It is based on 46 main indicators, organised in six chapters. The four central chapters divide the population by age into children (aged under 16); young adults (aged 16-24); adults (from 25 to 'normal retirement age'); and older people. The first chapter addresses poverty and low income, the last 'communities' (Tables 1&2). Within the chapters there are indicators relating to various themes, including health, education, and access to services as well as income. A summary table at the start of the report sets out all 46 indicators; whether they have improved or deteriorated over the past year, and since 1990; and the numbers of people affected in the latest year for which figures are available.
All of the indicators draw on existing, routinely collected data sets, although the authors list a number of topics which are omitted from their battery of indicators, because no regular and reliable data is available (Table 3). To qualify for inclusion as an indicator in the NPI report, the relevant data must be 'collected regularly and frequently' and 'must be reputable and generally accepted as a valid measure of the phenomenon being counted'. This, say the authors, 'creates a preference for official statistics' - although they note that this is neither a necessary nor a sufficient condition of high quality data (Howarth et al. 1998:13). Although the NPI does give a brief assessment of the quality of each indicator, there is little information about the background to the statistics. Readers of Radical Statistics will be well aware of the limitations of many official statistics as valid measures of the phenomena they purport to measure (Dorling and Simpson 1998, Irvine et al. 1979, Levitas and Guy 1996). However, when a package of indicators is drawn from multiple sources, as these necessarily are, criticism of individual indicators may easily be deflected on the grounds that they are only a small part of the overall picture. There may be an increased danger that the conditions of their production will be forgotten or ignored.
There is no clear definition of how social exclusion is distinct from poverty: 'Poverty and social exclusion are concerned with a lack of possessions, or an inability to do things, that are in some sense considered normal by society as a whole' (Howarth et al. 1998:18). Indeed, they appear to be treated as synonymous, since 'the notion of poverty that has guided the ... report is that where many people lack the opportunities that are available to the average citizen. ... This broad concept of poverty coincides with the emerging concept of social exclusion' (Howarth et al. 1998, p.13) - a view closer to RED than SID or MUD. However, on the next page, there is a shift into SID. A limitation of the report, in the authors' own terms, is 'a lack of clarity about what social exclusion might mean for a particular group', and thus how it might be measured. This is seen as a particular problem for older people, 'because neither inclusion within education and training nor inclusion within paid work will be central to overcoming any problem' (Howarth et al. 1998, p.14).
The indicators can be seen to include many items which are clearly relevant to RED, notably a battery of indicators of income poverty, as well as labour market attachment. But there are assumptions buried in some of the indicators, which may not be immediately obvious. 'Workless households' are widely presumed to be a problem. Among non-pensioner workless households, 21 per cent are headed by non-employed lone-parents. They are a problem for RED because they are likely to be poor, and children growing up in them are likely to be in poverty; for both SID and MUD they represent a problem of a different kind, since for SID they are 'excluded' by virtue of non-participation in paid labour, while for MUD they represent a moral hazard. The mere description of them as 'workless' ignores the unpaid work of lone parents and biases the interpretation towards SID or MUD rather than RED. Having divorced parents figures as an indicator for children - elaborated as an indicator of instability. But is having divorced parents an indicator of social exclusion? Or is it a risk factor in social exclusion? And if the latter, to what extent is this a consequence of the greater risk of poverty to which the children of divorced parents are exposed? The choice of indicators may also naturalise processes which are in reality outcomes of policy choices, and thus misrepresent causation. 'Pensioners with no private income' is an indicator of pensioner poverty only because the levels of the State pensions are so low. The policy implications of using this indicator (rather than pensioner poverty per se) is that the 'problem' is the absence of private pension provision rather than the inadequacy of collective provision.
The IPPR submission
The IPPR report, Social Exclusion Indicators, is a shorter document, produced in July 1998 as a submission to the Social Exclusion Unit before it abandoned work on this subject. This suggests a more compact index than the NPI report, with one lead indicator and some supplementary ones in each of four areas - income poverty, (un)employment, education and health (Table 4). The criteria for choosing indicators within these groups are that they should: be easily understood by the public and congruent with their concerns; be relatively easy to quantify; follow international conventions; have a dynamic dimension; and be able to be operationalised at the local area level. However, this is not because the view of social exclusion is narrower than that of the NPI. The four areas are reduced from an initial seven drawn from the concerns of the SEU itself: unemployment, poor skills, low incomes, poor housing, high crime environments, bad health and family breakdown. The complex causal relationships between these seven areas are briefly discussed, before reduction to four on the grounds of clarity and simplicity. The IPPR contains a more extended consideration of the strengths and weaknesses of its selected indicators than does the NPI. A central criterion for the key indicators is that 'they are capable of being clearly defined, measured and tracked' (Robinson and Oppenheim, 1998: 4); as so often reliability takes precedence over validity. But the final shortlist of indicators is seen as a first step: 'In the future, we hope further indicators will be developed to assess disadvantage from poor housing, high crime environments, family breakdown, and social and political exclusion, omitted from this report as they are difficult to extract from existing data sources. It is essential to develop indicators of social capital at a later date. Initial suggestions include the proportion of population who are members of a civic organisation and the extent of social support networks' (Robinson and Oppenheim 1998:ii).
The mention of social exclusion in this list is a little odd, but suggests that the key indicators are not, in fact, seen as indicators of social exclusion, even though they may be important to holding the government to account. The focus on 'harder' measures of poverty and inequality reflects a historical concern with such issues in British social policy, but also 'genuine difficulties in quantifying ... less tangible aspects of social exclusion' (Robinson and Oppenheim 1998:26). The authors argue that 'it is as yet unclear how one would define, measure and track social and political exclusion', and that consideration needs to be given to developing such measures. The very brief discussion of this suggests the possibility of using the British Household Panel Survey (BHPS) to measure 'social capital' by looking at data on social support networks, membership of civic organisations, and, for elderly people, possession of a telephone. Notably, two of these are included in the NPI's 46 indicators. Data from the Family Expenditure Survey shows 300,000 older people to be without a telephone. BHPS data cited in the NPI report shows that of the nearly 17 million adults not in full-time work or study, 9 million or 54 per cent 'do not participate in social, political or community organisations'. What this actually means is debatable, since the non-participation rate for other adults was 45%, and some forms of participation are work related. (The organisations listed are trade unions and professional organisations, parents' associations, pensioner groups, community and tenant groups, women's groups, religious groups, sports and social groups, and political parties).
Developing a consensual measure: Breadline Britain
There are serious limitations in starting from existing sets of statistics which have not been designed to measure social exclusion. Both the NPI and IPPR reports recognise that existing indicators are inadequate and do not really address the core question of what social exclusion is - and that new indicators need to be developed. An alternative (or at least supplementary) approach to working backwards from existing data sets is the extension of the Breadline Britain methodology explicitly to cover social exclusion (Gordon and Pantazis 1997, Bradshaw et al. 1998). This methodology seeks to establish a consensual view of poverty, through questions which not only explore the resources to which people (do not) have access, but also ask whether they regard them as necessities. In fact, because Breadline Britain was theoretically located in RED, and assumed that poverty results in exclusion from participation in social relationships, the original survey included questions about activities as well as consumption. It therefore provided not only a indication of some degree of social consensus about what people should expect to have as a minimal level of consumption, but what they should be able to do as a minimal level of social participation. The 1998 revisions to the original questionnaire extend its coverage of participation, not only in civic organisations, but in familial and friendship networks. The new schedule continues to collect information about multiple deprivation, including many of the topics included in the NPI list, but where participation is concerned it shifts the focus of attention away from institutional participation towards social networks. It begins to address the question of social exclusion directly, and the causes of it, thus addressing some of the issues noted in the NPI report as in need of further development.
The new Breadline Britain schedule shifts some of the power to define what kinds of participation are important from the political and social policy circuit to the broader public arena. Robinson and Oppenheim (1998) mention the need for indicators to be congruent with public concerns, and this is one of the strengths of the consensual approach of Breadline Britain. For example, people might now be asked if they thought all adults should be able to engage in a list of activities, including holidays, family celebrations, and occasional meals out - as well as whether they themselves are able to do these things, and if not, whether they are prevented from doing them by shortage of money, mobility problems or inflexible work commitments. It is possible, through this approach, to uncover how many people are excluded from aspects of sociality which are widely held to be, in currently unpopular terminology, a right to which citizens are entitled. One should not overstate the degree of 'democratisation' of the concept of social exclusion which can be achieved by this methodology. The consensual approach is used only for part of the questionnaire, and of course the categories offered to respondents, even if validated in focus groups, remain 'informed' or directed by the concerns of researchers. One should remember, too, that the consensual method may in part simply recycle popular prejudices, which may run counter to the prejudices of Ministers and policy wonks or may reflect them.
At the time of writing it is not clear whether and how the updated Breadline Britain survey will be carried out. If it proceeds, it will be the first large-scale attempt to address social exclusion other than through existing indicators developed for other purposes, and relate it to comparable data on income poverty, multiple deprivation and employment data. It will by no means solve all the problems of defining and measuring social exclusion. But it will be a first step away from the ad hoc compilation of tracks from the dodgy records of existing official statistics.
All the reports discussed here involve a battery of indicators, although the list proposed by the IPPR is somewhat shorter and more concentrated than that of the NPI. A comprehensive assault on social exclusion might be expected to produce an improvement on all indicators. There is, however, no reason to suppose that all the factors measured covary It is entirely possible that some will improve, some will deteriorate, some remain stable. In terms both of policies for addressing inclusion, and indicators for monitoring it, the priority given to different indicators, as well as the quality of the indicators themselves, is crucial. Even if a broad-ranging set of indicators is published including existing data sets on poverty, it is by no means clear that the question of inadequate income and resources will have the highest priority among these indicators, let alone that the adequacy of existing indicators in this and other areas will be questioned.
The question of priorities is crucial. Davey Smith et al. (1998) point out that the Independent Inquiry into Inequalities in Health (Acheson 1998), which was asked to make recommendations for policies which would reduce inequalities in health, makes 39 recommendations (with subsets bringing the total to 73). The recommendation to reduce poverty and inequality (especially by increasing benefits to women of child-bearing age, expectant mothers, young children and older people) has the same status 'as those regarding reducing traffic speed or offering concessionary fares to pensioners' (Davey Smith et al. 1998, p.1). The use of multiple (and unranked and uncosted) success measures seems to be becoming a habit. The March 1998 Green Paper New Ambitions for our Country: A New Contract for Welfare, published before the departure of both Harriet Harman and Frank Field from the Department of Social Security, also contained 32 criteria for judging the success of welfare reform - notably not including any explicit mention of a reduction in poverty (DSS 1998). The danger, as is pointed out in Our Healthier Nation, is that 'if everything is to be a priority then nothing will be a priority' (DoH 1998:57).
In the case of a putative report on poverty and exclusion, Howarth and Kenway (1998:85) see the non-ranking of separate indicators as a strength. Their broad range 'points up the fact that the report ... would not attempt to give a catch-all definition of poverty or social exclusion'; rather, 'each indicator appears independently instead of disappearing into an amalgam'. They argue that this gives flexibility, for it enables indicators to be added or dropped, and allows people to focus on what matters to them most. This begs the question 'which people'? The downside of these diverse packages of indicators is that they allow governments, as well as less powerful users, to pick and choose among the elements that interest them. And it is almost certainly the more powerful who will, in the end, determine which indicators are added or dropped. For what is selected or prioritised is not simply a matter of the interests of, say, Age Concern in the NPI's section on older people. It involves presumptions about the causal relationship between the social processes represented by the indicators, as well as their relative importance. NOT to prioritise indicators is to adopt an agnostic stance in relation to causality - and to leave the field clear for politicians, who are likely to plump for those which show them in the best light, because they are the cheapest to address, or have the greatest populist appeal, or are improving anyway for reasons unconnected with Government policy. The dangers of a minimalist approach, or of emphasising indicators which are more redolent of MUD than of SID or RED, are illustrated by the concerns of the Social Exclusion Unit itself.
The Social Exclusion Unit
The definition of social exclusion offered by the SEU is, as we have seen, imprecise. The meaning attached to it is perhaps better understood through the problems the Unit addresses. Its remit for 1998, besides the aborted project of developing indicators, was clear: to reduce truancy and school exclusions; to reduce rough sleeping; to develop 'integrated and sustainable approaches to the problems of the worst housing estates, including crime, drugs, unemployment, community breakdown, bad schools'; to consider 'preventive interventions with children and young people' and aspects of exclusion disproportionately affecting ethnic minority groups; and to consider ways of improving access to services for poor areas and poor individuals. The indicators selected in New Ambitions for our Country reflect this agenda. The principle that there should be specific action to tackle social exclusion and help those in poverty produces three criteria: a reduction in the scale of truancy and school exclusions; fewer people sleeping rough; and the introduction of a better model for tackling effectively the linked problems of the most deprived neighbourhoods (DSS 1998, p.84).
By the end of 1998, the SEU had issued three reports: Truancy and School Exclusion in May (1998a); Rough Sleeping in July (1998b); Bringing Britain Together: a national strategy for neighbourhood renewal in September (1998c). A fourth report, on reducing teenage pregnancies, was overdue. A consultation exercise took place in the autumn of 1998, but publication of the report is pending. A further consultation exercise is in place on policies for young people aged 16 to 19 who are in neither education nor training. There are strong elements of MUD in these concerns. The focus on truancy and school exclusions, especially given the overt linking of these to potential and actual criminality, together with the issue of teenage pregnancy, calls up the traditional demons of the 'dangerous classes': idle criminal young men and sexually/ reproductively delinquent young women. The focus on 'rough sleeping', rather than homelessness, may suggest a concern with social order as much as deprivation. Bringing Britain Together is less obviously focused on moral questions, although as Watt and Jacobs (1999) have shown, these are nevertheless present. This report does have stronger elements of RED, in making clear the necessity for the redirection of resources to deal with multiple deprivation; and elements of RED are present, too, in the earlier reports in so far as they propose better resourcing of particular services such as pupil referral units.
Some of these issues are picked up in the NPI report, where those 'permanently excluded from school' and 'births to girls conceiving under age 16' are included. These are narrower than the concerns of the SEU, which include truancy, and extend to pregnancy up to age 19, whether resulting in abortion or birth. The crucial question, however, is what priority is to be given to these indicators as representing 'social exclusion', by comparison with, for example, poverty? Moreover, given some other pronouncements by Ministers, there may be cause for concern over how 'social exclusion' is understood. In January 1999, Jack Straw suggested that teenage mothers should be encouraged to look on adoption as a positive option for their children. Lone parenthood - particularly young single parenthood - is typically seen as leading to social exclusion for both parent and child as a consequence of poverty, and for young mothers because of their detachment from education or training. Yet the suggested exclusion of such young women from their social role, obligation and status as mothers, and exclusion of children from their family of origin, is social exclusion of a profound kind.
There are three disadvantages to prioritising these as indicators: first, the danger of stigmatising the 'excluded' groups; second, the failure to interrogate WHY teenage pregnancy results in poverty and social exclusion (perhaps partly because of the poor level of benefits, as the Acheson report suggests); and third, the implied size of the problem. According to the NPI, there were 13,000 children permanently excluded from school in 1996/7, and in 1996 4,300 births to young women conceiving before they were 16. The SEU's figures on truancy suggest that up to a million children sometimes skip school, but this is not in itself necessarily significant in terms of social exclusion. More comprehensive figures on teenage pregnancy show that there were 8,800 (known) conceptions among under 16 year-olds, resulting in 4,500 abortions as well as the 4,300 births. Official figures are given for overlapping categories, and the comparable figures for those aged 15 to 19 were 94,400 known conceptions, resulting in 59,700 births and 34,700 abortions. These figures are tiny by comparison with the 3.3 million children living in households with below half average income in 1996/7, or the total of 10.5 million individuals of all ages living in such households. By focussing on particular social groups, the government can divert attention from the wider prevalence of social exclusion, from the link with poverty and from the causes of that poverty. Watt and Jacobs (1999) make a similar point about the focus on the 'worst estates' in Bringing Britain Together.
The inadequacies of current work on social exclusion - including lack of clarity as to causal processes, restriction to existing data which is not necessarily appropriate for the purpose, and lack of a sound basis for weighting indicators - have political implications. In particular, the opportunity for politicians to 'pick and mix' among indicators and among groups identified as socially excluded enables them to justify preferred policies and avoid confronting the growth of poverty and income inequality. Targetting particular social groups makes it possible to claim 'success' in reducing social exclusion without addressing the fundamental issues of poverty and inequality which afflict large parts of the population. We should be relieved that the task of developing indicators of social exclusion has moved from the SEU into a wider policy arena, but recognise that this is only the start of a struggle over what those indicators are to be.
Acheson, D. (1998), Independent Inquiry into Inequalities in Health, London: The Stationery Office
Bradshaw, J. and Gordon D., Levitas R., Middleton S., Pantazis C., Payne S., and Townsend T. (1998), Perceptions of Poverty and Social Exclusion: Report on Preparatory Research, Bristol: Townsend Centre for International Poverty Research, University of Bristol
Davey-Smith, G. Dorling, D., Gordon, D. and Shaw, M. (1998), The Widening Health Gap - what are the solutions? Bristol: Townsend Centre for International Poverty Research, University of Bristol
Department of Health (1998), Our Healthier Nation: a contract for health, Cm3852, London: The Stationery Office
Department of Social Security (1998), New Ambitions for our Country: A New Contract for Welfare, Cm 3805, London: The Stationery Office
Dorling, D. and Simpson, L. (1998), Statistics in Society, London: Arnold
Duffy, K. (1995), Social Exclusion and Human Dignity in Europe, Strasbourg: Council of Europe
Gordon, D. and Pantazis, C. (eds.) (1997), Breadline Britain in the 1990s, Aldershot: Ashgate
Howarth, C. and Kenway, P. (1998), 'A multi-dimensional approach to social exclusion indicators' in Carey Oppenheim (ed.) An Inclusive Society: Strategies for tackling poverty, London: IPPR
Howarth, C. and Kenway, P., Palmer, G. and Street, C. (1998), Monitoring Poverty and Social Exclusion: Labour's Inheritance, York: New Policy Institute/Joseph Rowntree Foundation
Irvine, J., Miles, I. and Evans, J. (1997), Demystifying Official Statistics, London: Pluto Press
Levitas, R. (1998), The Inclusive Society: Social Exclusion and New Labour, Basingstoke: Macmillan
Robinson, P. and Oppenheim, C. (1998), Social Exclusion Indicators: A submission to the Social Exclusion Unit, London: IPPR
Levitas, R. and Guy, W. (eds.) (1996), Interpreting Official Statistics, London: Routledge
Morris, L. (1994), Dangerous Classes: The Underclass and Social Citizenship, London: Routledge
Murray, C. (1990), The Emerging British Underclass, London: Institute of Economic Affairs
Murray, C. (1999), The Underclass Revisited, American Enterprise Institute for Public Policy Research (see http://www.aei.org/ps/psmurray.htm)
Social Exclusion Unit (1997), Social Exclusion Unit: Purpose, work priorities and working methods, London: The Stationery Office
Social Exclusion Unit (1998a), Truancy and School Exclusion, London: The Stationery Office
Social Exclusion Unit (1998b), Rough Sleeping, Cm 4008, London: The Stationery Office
Social Exclusion Unit (1998c), Bringing Britain Together: a national strategy for neighbourhood renewal, Cm 4045, London: The Stationery Office
Townsend, P. (1979), Poverty in the United Kingdom, Harmondsworth: Penguin
United Nations (1995), The Copenhagen Declaration and Programme of Action: World Summit for Social Development 6-12 March 1995, New York: United Nations Department of Publications
Walker, A. and Walker, C. (eds.)(1997), Britain Divided: The Growth of Social Exclusion in the 1980s and 1990s, London: Child Poverty Action Group
Watt, P. and Jacobs, K. (1999), 'Discourses of Social Exclusion: An Analysis of Bringing Britain Together: a national strategy for neighbourhood renewal', paper presented to Discourse and Policy Change Conference, University of Glasgow, February 1999