Social exclusion and multiplicity

Poverty, to paraphrase the bible, has always been with us, but it is conceptualised and measured in different ways at different times. It is broadly recognised that single measures, such as income poverty, are inadequate to capture the effects and extent of poverty. Peter Saunders (Reference Saunders2005) provides an account of the Australian ‘poverty wars’ that led to research and policy moving from a focus on the poverty line to a focus on social exclusion and deprivation. Jacqueline Best (Reference Best2013) traces changes in the way that the World Bank represents global poverty, shifting from a view of structural adjustment to one of risks and vulnerabilities. In contrast to an emphasis on policy shocks, such as the global financial crisis, many contemporary accounts of poverty focus on multiple disadvantages as both cause and effect: mental health problems, long-term joblessness and alcohol and drug misuse recur as examples.

Debates around the poverty line are always technical and political. There are also debates around multiple disadvantages, such as arguments about their cause and nature, and how best to address them. Welshman's (Reference Welshman2007) history of poverty debates traces the contests between agency-based accounts and structure-based accounts of poverty back to Titmuss. There is relatively little argument, however, about the importance of multiple exclusions or risks, which finds consensus in the epidemiological and sociological literature, and in policy discourse. A notable exception to this consensus is the ‘troubled families’ argument, to be detailed later.

The concept of ‘social exclusion’ is one example of this consensus. In both the UK and Australia, it emerged from research and advocacy, and was taken up formally in government agendas and strategies. In part a recognition that socioeconomic disadvantage has many dimensions, social exclusion is also described in terms of individual experiences: ‘multiple disadvantage can have a compounding and persistent effect, reinforcing barriers to getting ahead and increasing the likelihood of other related problems later in life’ (Australian Social Inclusion Board, 2012: 6).

A different strand of research also highlights the importance of multiple disadvantages in determining individual and intergenerational outcomes. Large-scale longitudinal studies which study the prevalence and effects of specific characteristics or ‘risk factors’ show that no one single risk is as important as the number of risks to which people are exposed (Danese et al., Reference Danese, Moffitt, Harrington, Milne, Polanczyk, Pariante, Poulton and Caspi2009; MacKenzie et al., Reference MacKenzie, Kotch and Lee2011). In this, social policy research mirrors chronic disease research, where, as Armstrong (Reference Armstrong2012) notes, a focus on screening as a preventive measure against diseases (viz tuberculosis in the early twentieth century) has been displaced by attention on multifactorial causation and risk factors.

Moreover, problems are not only multiple, they are multiply constituted. Consider, for example, child maltreatment, which has a prominent place in most accounts of disadvantage as both a cause and effect of social exclusion: children from multiply disadvantaged families are at a much higher risk of maltreatment than others, and children and adults who have experienced maltreatment are at a much higher risk of experiencing multiple disadvantage than those who have not. Child maltreatment is an example of over-determination or equifinality, that is, multiple causal risk pathways lead to one outcome (MacKenzie et al., Reference MacKenzie, Kotch and Lee2011). Historically, the family was the unit of analysis for maltreatment research, but over time this focus has been expanded. Since the 1970s, ecological frameworks have been directing attention outwards from the individual, in recognising the importance of social environments. More recently, epigenetic research points inwards to within individuals, and the interactions between genetic and social domains.

The risks for child maltreatment therefore reside in multiple domains, and this is also true of other phenomena. A fairly parsimonious list of risk factors (MacKenzie et al., Reference MacKenzie, Kotch and Lee2011) includes ten variables; a meta-analysis (Stith et al., Reference Stith, Liu, Davies, Boykin, Alder, Harris, Som, McPherson and Dees2009) includes thirty-nine. These variables include characteristics of the parent, the child, the child–parent relationship, the household and the neighbourhood. Although more difficult to measure at the level of individual outcomes, national characteristics, including early education, parental leave and the social gradient, are also recognised as important to the safety of children in families (Gilbert et al., Reference Gilbert, Fluke, O'Donnell, Gonzalez-Izquierdo, Brownell, Gulliver, Janson and Sidebotham2012).

This has implications also for the multiplicity of domains in which social policy is now accountable. Ecological approaches construct a temporal dimension to social policy through the promotion of early intervention and prevention. In the fields of mental health, homelessness, domestic and family violence, crime and child maltreatment, policies are designed to address warning signs early to stop them escalating. Rather than viewing social exclusion as an individual problem, and the responsibility of intensive services who treat them, these models look to an array of risks, and place responsibility for preventing and intervening in the community and in generalist or universal services, including schools, public health systems and childcare.

The conceptual basis of contemporary social policy is therefore one of multiplicity. Social disadvantage is caused by, and manifested in, multiple problems; those problems have multiple origins; and there are multiple locations in which those people and problems can be identified and addressed.

This shift from poverty to social exclusion and multiple disadvantages is based on recognition of the multiple dimensions of disadvantage. In at least some accounts, social exclusion also has a structural force: if people are excluded, there are forces that are excluding them. But, as described in the following section, this is complicated by the evolution of a category of people, ‘multi-problem families’ who emerged in the 1950s, even before the ‘war on poverty’ was announced in 1964. Whereas the concept of social exclusion describes the social processes that exclude, the concept of multi-problem families describes the characteristics of people who are excluded. Recent developments in the concept of multi-problem families represent a convergence of these concepts, with the result that explanations of social processes dovetail into descriptions of individuals and families.

From multi-problem families to complex needs

Multi-problem families appear in the academic literature from the early 1960s. These families are rarely defined, but are described even in the early literature as a type with whom services are familiar: chaotic lives; co-occurring problematic substance use and mental health problems; generally living in poverty. Cunningham Dax and Hagger (Reference Cunningham Dax and Hagger1977: 227) are lurid: ‘homes are usually neglected and unhygienic, there is little household routine, meals are unpredictable and irregular. Fights between intoxicated parents are commonly witnessed by the young. Contacts with the welfare departments and the law are frequent and one or another member is often in prison.’ Trevvett (Reference Trevvett1967: 308) laments that increased spending on services for multi-problem families has not had the outcomes it should: ‘their numbers are increasing and their problems are becoming more serious’.

Another key characteristic of multi-problem families, according to this early literature, is the intensity of their involvement with services. Curtis et al. (Reference Curtis, Simon, Boykin and Noe1963: 510) open their article by noting the disproportionate impact of multi-problem families on services: making up only 6 per cent of families but accounting for two thirds of families receiving correctional, health and mental health services and ‘public relief’. Trevvett (Reference Trevvett1967: 308) similarly notes that 80 per cent of agency caseload is taken up by 6 per cent of the population, while Cunningham Dax and Hagger (Reference Cunningham Dax and Hagger1977: 309) put the figure at 5 per cent. The origin of this figure is a study by Buell and colleagues in 1948, who studied data on the problems presented by families to support agencies in a city in Minnesota, and found that 6,600 families, or around 6 per cent, of the city's families were absorbing over half of the combined services of the agencies (Buell, Reference Buell1952: 9). In the UK in the 1970s, a different percentage was nominated by Keith Joseph, Secretary of State for Social Services, who was concerned with the problem of ‘transmitted deprivation’. He reported to a newspaper that he had been told by a service director: ‘we have 20,000 households in this city. Nearly all our problems – delinquency, truancy, deprivation, poverty and the rest – come from about 800 of them’ (cited in Welshman, Reference Welshman2007: 28). There have been few other attempts to count multi-problem families, partly because their characteristics are varied, and partly because almost by definition they escape the categories of most data collection exercises. Figures on mental health prevalence will include multi-problem families, but mental health data generally do not also indicate other family characteristics, such as insecure housing or family violence. Moreover, while mental health problems recur as important, not every multi-problem family has members with them.

Over time, the service involvement of multi-problem families has become more important to their conceptualisation, as research and service delivery has evolved from a focus on individual pathology and blaming families as ‘the problem’, to ecological and resource-based approaches (Sousa et al., Reference Sousa, Ribeiro and Rodrigues2007). Family systems therapy and research is still concerned with multi-problem families (Friedlander et al., Reference Friedlander, Lambert and de la Pena2008; Al et al., Reference Al, Stams, Bek, Damen, Asscher and van der Laan2012), but in the broader human services there is evidence of a shift to a focus on ‘complex needs’. In contrast to the term ‘multi-problem families’, which occurs in one or two articles a year, ‘complex needs’ has increased over time. Web of Science results show the phrase occurring once in articles published in 1990, growing to eighteen occurrences by 2000 and sixty-five in 2013. In the early 1990s, the term was used as a general descriptor: ‘the complex needs of youths today’ (Bennett and Westera, Reference Bennett and Westera1994); ‘the varied and often complex needs of families’ (Dakamulwanda et al., Reference Dakamulwanda, Thornburg, Filbert and Klein1995). By 2014, however, it is being used to describe groups of especially disadvantaged people, or those who present challenges to services: for example, young people with disabilities transitioning from out-of-home care (Mendes and Snow, Reference Mendes and Snow2014); and women with histories of childhood sexual abuse and/or domestic violence (Salter and Breckenridge, Reference Salter and Breckenridge2014).

This development of the concept of complex needs, from a generic descriptor of everyday life to a category of person, was exemplified by a 2004 report from the UK Institute for Public Policy Research (IPPR). The report described problems in the effective provision of services and support, organised under the rubric of ‘complex needs’. The report offers a straightforward gloss on people with complex needs: they are people with multiple interconnected needs that span medical and social issues (Rankin and Regan, Reference Rankin and Regan2004: 7). The report also rejects definitions of the term in favour of vignettes and case examples on the basis that ‘there is no typical complex needs case’ and ‘a definitive label is neither meaningful nor useful’ (Rankin and Regan, Reference Rankin and Regan2004: 7). Somewhat in tension with this inclusive framework, it also notes that systematic data should be collected on service use to produce more accurate counting of people with complex needs; in the absence of these data, the ‘component parts’, which are relatively well-documented, are enumerated. These include contact with specialist mental health services; severe/profound learning difficulties; class A drug users; people with alcohol problems or addiction; and children permanently excluded from school (Rankin and Regan, Reference Rankin and Regan2004: 20).

Although arguing that it should not be seen as ‘the latest label’, the report is nonetheless explicit in its advocacy of ‘complex needs’ as a class of people to whom services and policy should be responding:

The IPPR report was not the first to use the term ‘complex needs’, but clearly creates it as a category. It brings into being a new class of people who have particular characteristics and similarities with each other, and advocates for their recognition by services and policy as a class of people. This new class of people has similarities to those described in other systems, particularly those oriented around service delivery, such as ‘hard to reach’ or ‘vulnerable’; what distinguishes this classification is the importance of multiple vulnerabilities, which intersect with each other and have a multiplying effect. This is an attempt at what Hacking (Reference Hacking2002a, b) calls ‘making up people’, or the process by which classifying people creates new and specific ways of being in the world: ‘classifications that, when known by people or by those around them, and put to work in institutions, change the ways in which individuals experience themselves’ (Hacking, Reference Hacking2002b: 11). Categories work through institutions and individuals, and make certain experiences and responses possible while foreclosing others. The best-known examples of making up people are the categories of the ‘homosexual’ and the ‘schizophrenic’. The invention of these categories did not lead to practices and sensibilities, but they did bring about changes: to medical, judicial and political interventions, to social identities, and to how people understand themselves and others.

Best (Reference Best2013) argues that shifts in the ways that global poverty is conceptualised is an example of making up people. Social protection policies seek to promote desirable practices and prevent undesirable ones; labour market policies no longer focus only on barriers to labour market flexibility but also work to produce a better trained, ‘work ready’ population. This is a ‘new way of conceptualising the poor, together with [a] whole wealth of new strategies designed to make this measurement, representation and management possible’ (Best, Reference Best2013: 122).

We can see, similarly, that shifts in conceptualisations of poverty in rich countries also exemplify making up people. Social exclusion is one instance of this, and delineates the prevalence of multiple forms of disadvantage and the associations between them. ‘Complex needs’ is another instance, based on the characteristics and service needs of individuals and families. As noted earlier, both are accounts of poverty in which multiplicity is central.

But this is an even more complicated story. Statistics are essential to the classification of people, and to the organisation of practices into prevalence data. There are serious attempts to count the ‘deeply excluded’, just as there have long been serious attempts to count the poor. In the case of ‘complex needs’, however, statistics find their greatest use in the epidemiological studies cited earlier, that is, in manipulating data to understand the prevalence and impact of different characteristics in specific populations. We have never known how many multi-problem families there are, we do not know how many people live with complex needs.

This seems an odd reversal to the usual trends. Population statistics are generally getting much better. If we haven't got better at counting and defining complex needs, what has happened instead? To answer this, we can turn to another instance of the rise of multiplicity over the last forty years: in the conceptual and disciplinary approaches that constitute social policy research. A shift in focus from the groups of people being described, to the disciplines and techniques that describe them, illuminates the historical development of the conceptualisation of poverty as multiple and complex needs.

Social policy and multi-disciplinarily

Forty-five years ago Donald Campbell's article, ‘Reforms as experiments’ (Reference Campbell1969), inaugurated our contemporary era of policy as self-consciously experimental, and signalled the emergence of what is now known as evidence-based policy. Campbell argued that rigorous evaluations of policies were rarely carried out, because the careers of administrators and politicians were tied to the fate of those policies. This, he argued made for ineffective responses to social problems, and instead of advocating a particular policy, governments should take seriously the importance of addressing the problems. This would mean a shift in the governance of social problems:

In addition to this advocacy, the article also described the results of empirical research on drunk driving, specifically the effect of changes to the British Road Safety Act which introduced breathalysing, mandatory punishments for driving over the limit and strong publicity about the specific starting date for new regulations. The analysis showed a decline in casualties and serious fatalities as a result of the crackdown, but the article's primary concern is the methodology by which these findings were derived, and the alternative hypotheses considered other than the crackdown that could have caused this change. He considered and rejected other hypotheses, including publicity of the crackdown, improvements in traffic control, fewer motorbikes (which have more accidents than cars) and incentives from insurance companies for not having accidents. The method by which Campbell derived his findings, interrupted time series analysis, is still used, although is not regarded as providing the rigour of other methods, most obviously randomised control trials (RCTs).

Note that in this article Campbell was not especially concerned with understanding human behaviour, or even concerned that it may be difficult to do so. There is no sign of complex problems or people with multiple needs here. The task that Campbell and colleagues identified for themselves was not in understanding people or problems, but in making sure that what was claimed to be changing behaviours, actually was changing behaviours.

Since that time, experimental approaches have become established as an important, although by no means the only, way of implementing and measuring social policy change. The Campbell Collaboration, named after Donald Campbell, is modelled on the Cochrane Collaboration as a database of systematic reviews in education, crime and justice, social welfare and international development (http://www.campbellcollaboration.org/). The human services have models of trials and evaluation derived from clinical trials, and evidence-based policy is commonly described as a goal or principle of governments and policy agencies.

Yet Campbell's aim of trialling policies on an experimental basis, testing to see if they are effective, and abandoning them if they don't work for a new approach, remains distant. Although Campbell advocated a political stance of technical experimentation, each of the key elements of his statement – initiating a policy, defining improvement, measuring to evaluate improvement – turn out to be more complicated than that. For the Campbell of the 1969 article, robust measurement of social practices was a complex task, but the practices themselves were not especially complex. In the contemporary moment, robust measurement of social practices is an enormously complex task and the practices themselves are complex.

Four years after Campbell's article was published, Rittel and Webber's ‘Dilemmas in a general theory of planning’ (Reference Rittel and Webber1973) was another inaugurating moment. The article introduced the concept of wicked problems, which are contested, messy and provisionally solved, and their differences from non-wicked problems, which are stable, definable and separable, and generally found in the physical sciences. Wicked problems are characterised by uncertainty, emergence and contest. They have no definitive formulation, and the different ways they are defined suggest different solutions. They have multiple interdependencies, no clear beginning and no absolute solution, and every problem can also be seen as a symptom of another problem. Paradigmatic examples of contemporary wicked problems are child protection, climate change and ‘self-harming’ behaviours such as smoking and unhealthy eating (Head, Reference Head2008).

It is possible to see the contemporary social policy landscape as the legacy of both Campbell, and of Ritter and Webber. Hierarchies of evidence are important to policy, and experimental approaches to policy design and measurement are privileged, if only in the sense that they are more highly valued than other approaches. It would be misleading, however, to present the uptake of research in policy as monolithic, or the RCT model as all-conquering.

Instead, the conceptual foundations of policy include multiple accounts of social practices and behaviour, and the effects of policy are measured in multiple ways. On the one hand, Campbell's ghost can be seen in the rise of evidence-based policy and practice, the growing influence of expertise and professionalism in the human services and the centrality of evaluation and monitoring to policy. It can be seen in the quantification and measurement of populations in RCTS and longitudinal studies. This is the engineering or bridge-building model of policy research, where social science provides technical answers to help solve policy problems (Bulmer, Reference Bulmer1982).

On the other hand, ‘wicked’ problems are an established trope in social policy, where the political, contested and normative nature of policy formation is central to both design and implementation, and experimental approaches to evaluation are considered alongside realist, developmental and post-positivist approaches (Pawson and Tilley, Reference Pawson and Tilley1997; Patton, Reference Patton2010; Wertz et al., Reference Wertz, Charmaz, McMullen, Josselson, Anderson and McSpadden2011). If this is policy research as engineering, it is the engineering described in Langdon Winner's ‘Do artifacts have politics?’ (Reference Winner1980), in which technologies are inherently political, and so too is knowledge: insights which are also clear in critiques of evidence-based policy (Marston and Watts, Reference Marston and Watts2003; Jacobs and Manzi, Reference Jacobs and Manzi2013). This is evident in the conceptual shift from multiple problems to complex needs. As noted, this is a shift from focusing on individuals as dysfunctional to a focus on their needs, and, implicitly, the responsibility of services to meet them. This shift could not have happened without the influence of social movements and activism on social policy. There are inexact numbers (at best) on those with ‘complex needs’, because the concept has a different function from that served by RCTs. The quantification of characteristics, risks and intervention effects is one important part of social policy research. The concept of complex needs contributes to another important part concerned with the intersections of different values, systems and practices.

Since Campbell's article was published in 1969, policy understandings of human behaviour have become more sophisticated and complete. These changes have happened in part because of the influence on policy from a range of disciplines and scholarly fields. Economics and psychology are the most dominant of these, but others including anthropology, sociology, historical studies and post-colonial studies are also important. New public management emerged from business studies, evidence-based practice from clinical trials, risk factor research from epidemiology. But policy is also, primarily, constituted by political advocacy and political contest, and these multidisciplinary approaches are only another dimension of that inherent multiplicity. We can see that exemplified in our final example of multiplicity: the invention of ‘troubled families’.

Contesting categories: ‘troubled families’

In 2011, seven years after the publication of the IPPR report, people with complex needs took on a more menacing policy hue. After nearly a week of ‘riots’ across several London boroughs and towns across England following a police shooting, the UK government announced a series of initiatives to target 120,000 ‘troubled families’ across England. Prime Minister David Cameron's speech makes an indirect link between those with ‘complex needs’ and those who are ‘troubled’, although the phrasing here is ‘multiple disadvantages’, and the language is closer to that used to describe multi-problem families in the 1970s:

This identification of ‘troubled families’ received a critical reception, along two primary axes. First, and more familiarly, blaming social problems on the behaviours of disadvantaged people was criticised as individualising structural problems that are in fact the responsibility of governments. This was one instance of a longstanding critical tradition, in which accounts of poverty based on social agency are contested by materialist accounts (Deacon, Reference Deacon2004; Welshman, Reference Welshman2007). Second, and unusually, the number of families was disputed. Ruth Levitas (Reference Levitas2012) called the ‘imputed existence’ of 120,000 families a ‘factoid’ and argued the term ‘deliberately conflates families experiencing multiple disadvantage and families that cause trouble’ (2012: 12). Jonathan Portes (Reference Portes2012) wrote that Cameron's speech ‘insulted, on the basis of an analysis that is not just flawed but plain wrong, a large number of families who have more than enough problems of their own’.

The basis of these criticisms was the unacknowledged shift from one definition of troubled families to another. The criteria for inclusion in the initiatives were narrow and specific: involvement in youth crime or anti-social behaviour; children regularly truanting or not in school; at least one adult on out of work benefits; causing high costs to the taxpayer. However, the estimated number of such families was derived from broader social exclusion data not related to crime: no parent in work; poor quality housing; no parent with qualifications; mother with mental health problems; one parent with longstanding disability/illness; family has low income; family cannot afford some food/clothing items.

The definitional slippage between the socially excluded and troubled families was recognised for its political effects of not merely blaming the poor for social discord, but conflating the behaviour of the poor and a tiny minority of people. However, and this was not recognised, the move was also a convergence of different ways of conceptualising and counting, a moment when two different meanings of multiplicity – social exclusion and complex needs – converged. Criminal activity, for example, is not included in the social exclusion indicators. However, it is a recurring indicator in studies of multiple disadvantage and complex needs (Nikulina et al., Reference Nikulina, Widom and Czaja2011). The Department for Communities and Local Government document announcing the Troubled Families programme noted also that ‘troubled families’ were characterised ‘almost always’ as having ‘long-standing problems which can lead to their children repeating the cycle of disadvantage’, including child protection concerns, mental and physical health problems, and isolation (Department for Communities and Local Government, 2012: 1). Troubled families, in other words, closely resemble the multi-problem problems of an earlier era, those with complex needs, and (via attempts to enumerate them) the socially excluded.

The contesting of the ‘troubled families’ category is unusual for social policy, and unusual historically: neither multi-problem families nor complex needs received the same criticism. It happened because different concepts of multiplicity as a way of describing people and multiple ways of conceptualising policy converged:

• • first, the data on social exclusion, which provide more robust data on the prevalence of particular types of disadvantage than any other, were used to define and count a population;

• • second, the familiar figures of multi-problem families/people with complex needs, were used to describe this population. A key descriptor here is that of disproportionate impact; that is, there are relatively few of these families, but they cause most of the problems and use most of the services;

• • third, political and conceptual debates around framing and measuring disadvantage were reanimated by the category. The ‘troubled families’ initiative attributed problems squarely to families and placed faith in the capacity of programs to correct them in quantifiable ways. Objections to the initiative were based partly on defending the behaviour of the poor, and partly on the attribution of social problems to that behaviour, rather than to the failure of government to address social exclusion;

• • fourth, the fact that ‘troubled families’ were brought into being at the same moment as the program to ‘turn their lives around’ typified the adage that policy does not solve pre-existing problems but rather constitutes both problems and solutions. More specifically, it exemplifies evidence-based policy and the concomitant importance of programmatic, experimental approaches to social interventions. The Troubled Families program has specific, measurable outcomes, which are published on the Department for Communities and Local Government website and are the basis on which the agencies providing the services are evaluated and paid (Department for Communities and Local Government, 2014).