Payne, G., J. Payne and M. Hyde (1996) '"Refuse of All Classes"? Social Indicators and Social Deprivation', Sociological Research Online, vol. 1, no. 1, <>.

Copyright Sociological Research Online, 1996


'Refuse of All Classes'? Social Indicators and Social Deprivation

by Geoff Payne, Judy Payne, and Mark Hyde
Faculty of Human Sciences at the University of Plymouth, UK

Received: 15/1/96      Accepted: 4/3/96      Published: 29/3/96


The development and electronic accessibility of indices of poverty and social deprivation have yet to be fully exploited by mainstream sociology, not least in the field of class analysis where it might seem likely to be taken up. While reasons for this can be suggested, there are several conceptual frameworks within sociological debates about class that might accommodate deprivation and its indicators, and also valuable empirical resources in the form of indices which are now available to researchers interested in contemporary social inequality.

The potential of this approach in the UK is demonstrated by an examination of patterns of social deprivation in 1991 Census data for 391 wards in the South West of England, using the Townsend, Jarman, Breadline Britain and the new DoE Local Conditions indices. Urban and rural patterns are demonstrated in inter-pair correlations between index scores, component variable values, and social class represented as SEGs. A factor analysis similarly shows distinct patterns for urban areas, small towns and rural areas. However, in all cases class, single-parent families, and children living in low- income households show the strongest associations with other deprivation indicators.

An explanation for the empirical findings may be found in two main strands of class analysis. First, following Weber, deprivation and occupational class both derive from market situations, but the reported deprivation patterns cannot be entirely explained in terms of class: other factors (such as life-cycle) need to be included. Second, while there is no clear evidence of residualization in the data, some aspects of consumption sector theory seem to be born out; for example, differential opportunities for access to consumption. In addition, it is suggested that the rural/urban differences raise issues for ameliorative policies, further demonstrating the potential for a closer integration of the social indicators approach into the techniques of sociological analysis.

Deprivation; Class; Poverty; Social Indicators; South West; Urban; Rural


The accumulation of new measurements of poverty and social deprivation is creating a growing resource for sociological analysis. At its simplest, we are now able to describe more precisely the nature and distribution of deprivation among the most disadvantaged members of society. This in turn provides a basis for exploring the inter-connection between dimensions of poverty, and the identification of typological patterns. At a more conceptual level, we can begin to locate types of deprivation within more established debates in class analysis where, paradoxically, there has been less developmental synthesis than might have been expected, given the amount of social research effort already devoted to social disadvantage. The potential of this approach can be shown by using electronically- available data from several indices of deprivation, and in particular, the new Index of Local Conditions (DoE, 1995).

Class And Deprivation

Several theories of social class offer a basis for an explanation of social deprivation. The Marxist tradition contains two main ways of thinking about this issue. First, Marx's observations about the lumpenproletariat imply that all societies contain a lower stratum of politically dangerous undesirables who are both deprived and depraved. He characterises the lumpenproletariat as an urban mass of

...thieves and criminals of all kinds, living on the crumbs of society, people without a definite trade, vagabonds, gens sans feu et sans aveu.....discharged soldiers, discharged jailbirds, escaped galley slaves, swindlers, mountebanks, lazzaroni, pickpockets, tricksters, gamblers, maquereaus, brothel keepers, porters, literati, organ-grinders, ragpickers, knife grinders, tinkers, beggars....this scum, offal, refuse of all classes. (Marx, 1962: p. 295)
This is an inadequate basis for analysis, and indeed, such is Marx's moral contempt and political loathing for this mass, that he says relatively little about it in his main works. Clearly deprivation in contemporary society cannot be characterized as applying only to the categories of person that Marx lists, nor as we shall see is deprivation a uniquely urban phenomenon. It is ironic that it has been the New Right that has rediscovered the denizens of the inner city and large housing schemes as an 'underclass' of dependent - and dangerous - undeserving poor, a characterization challenged by a number of sociologists (e.g. Wilson, 1987; Murray, 1990; Jencks and Peterson, 1991; Westergaard, 1992; Gallie, 1994).

A more sophisticated, if less specific, framework can be found in Marx's ideas of exploitation under capitalist production (Novak, 1988; George and Howards, 1991; Wright, 1995). Labour exploitation is premised on the exclusion of large numbers of the population from ownership of productive assets. The tendency of exploitation to create deprivation among those in employment is buttressed by the existence of 'reserve armies of labour', constituted by the non-working poor. Thus divisions within the working class and associated patterns of deprivation are an outcome of the way that capitalist labour markets are regulated (Novak, 1988; Mann, 1991). This includes the role of the welfare state as a mechanism to compensate for market failure, without altering its fundamental character (Titmuss, 1958; Marshall and Bottomore, 1992).

While this second Marxist perspective offers a general framework to account for poverty, its main thrust is towards economic exploitation as a system. It is less concerned with plotting the variety of differences in life-chances, and has offered little in terms of empirical operationalization and measurement. Nor can we use it as a handle on detailed patterns of deprivation: for instance, are these contingent, or is there a specific purposive need for capitalism to maintain a reserve army of, say, unemployed, low-skilled female workers in a given location? Why are lone-parent households so prone to poverty, and why do their numbers vary over time? It may be possible to make the necessary connections, but these have yet to be spelt out, given the level of analysis that has dominated Marxist accounts, and therefore this approach flatters to deceive.

More recently, several Marxist analyses have drawn attention to the consequences of trans-national corporations transferring manufacturing investment away from the UK to the countries of the 'Pacific rim', leading in turn to de-industrialization and rising unemployment in the UK (Brown and Scase, 1991; Brown and Crompton, 1994). Although unemployment has declined in the last few years, this has not translated into the kinds of employment opportunities that characterized the post-war labour market. UK-based employers have promoted non- standard work which now accounts for more than one-third of labour market opportunities (Brown and Scase, 1991; European Commission, 1995). These changes were accompanied by economic and labour market policies which, informed by 'supply side' economics, have aimed to reinforce work incentives by promoting 'workfare' and 'less eligibility' (Gaffikin and Morrissey, 1992; Townsend, 1993a).

One version of this approach, regulation theory, emphasises the way that capitalist societies are regulated around specific 'production norms' (Lipietz, 1992; Amin, 1993). Driven by economic crisis and industrial conflict, the full male employment and corporatist bargaining structures of 'Fordism' have given way to the labour market flexibility and welfare austerity of 'post-Fordism' (Jessop, 1993). For obvious reasons, intensified deprivation is an integral feature of this new 'mode of regulation'.

The main strength of this account is also a source of weakness as far as explaining the problem of deprivation is concerned. Like other 'Marxist' accounts, the regulation approach is:

...overly economically deterministic, allowing little space for the power of human agency and political struggle in patterns of social change. (Burrows and Loader, 1994: p. 4).
This focus on 'economic restructuring' has the capacity to explain deprivation that has labour market origins, but fails to incorporate factors that cannot be 'reduced' to productive relations. Thus the influence of demographic processes, such as changing family patterns and political decision-making of the kind that introduced the 'right to buy' housing policy, rarely feature in post-Fordist accounts of welfare.

As the main alternative, Weberian approaches offer several potential advantages. Following Lockwood, 'class position' includes both market and work situations, so that people can be conceived of as being poor because they or their partners can only obtain limited, or even no, paid employment in the labour market, i.e. work commanding low income and 'status'. Status here includes stigmatizing low status that reinforces the inherent disadvantage of the socially deprived who, because of inherited disadvantage, restricted opportunity networks, or discrimination, lack the human capital required for jobs with higher status and income (Le Grand, 1982; Pahl, 1984,1988; Payne, G., 1987; Goldthorpe and Marshall, 1992; but also see Payne and Ford, 1983). A new division in contemporary society is between the two-thirds in employment and the one-third unemployed, with distinctive life-styles and consumption patterns.

Although based in market situation, this shifts the emphasis into distributional outcomes and consumption sector theory. As Dunleavy argues, when a growing section of working class households have opted out of state provision of housing, transport, education and health-care, they perceive their interests in terms of their private consumption location and not in terms of the manual occupation of the main wage-earner. They come to believe that it is unfair to be taxed for services that they do not use. This leads to a realignment of voting behaviour around consumption sectors (Dunleavy, 1980; Mann, 1991). Saunders claims that the growth in the real incomes of working-class households has provided the necessary condition for privatization (for without it, there would be little demand). Second, the 'fiscal crisis of the state', as state expenditures outstrip state revenues, provides incentives for programmes of privatization. This de-collectivization of consumption leaves a minority dependent on a declining and increasingly residual public sector (Saunders, 1981). More recent formulations present this process as one in which groups are exposed to differential 'risks' of being unable to access collective goods (Beck, 1992) and, as Eder (1993) argues, compete in specific cultural or 'communications' spaces to manipulate meanings and values in new ways in order to form a basis for collective action.

The consumption-sector approach and subsequent elaborations allow a role for the working class in the production of inequalities. Insofar as they have dealt with this, Marxists have tended to highlight a lack of solidarity engendered by false consciousness, which works against an otherwise intrinsic tendency to support equality and universalism. In the consumption-sector approach, sections of the working class have played a major part in promoting inequalities by actively pursuing and participating in private consumption, albeit with assistance of the private sector and central government. This approach also avoids class reductionism, or the tendency to reduce all manifestations of inequality directly to divisions in the production process or the interests of capital, by acknowledging the opportunities that individuals have for improving their life-chances by utilizing resources outside of the production system. Third, it recognizes the potential of the welfare system to be an important source of inequality in its own right. In this way, it helps to shift the focus from The Class Struggles In France 1848-1850 to social exclusion in contemporary Britain (DSS, 1992; Townsend, 1993b; Green, 1994; Power and Tunstall, 1995; and the report of the Joseph Rowntree Foundation's Inquiry into Income and Wealth, 1995).

Social Indicators And Social Class

Despite this choice of analytical frameworks, such recent research findings have yet to be well integrated into mainstream sociology. In part this is due to the academic division of labour: deprivation 'belongs' to the applied and quantitative areas of social policy (Payne, J., 1995a) whereas social class 'belongs' to the more theoretical field of sociology. The potential for sociology of deprivation indices, such as the Jarman Index (1983), the DoE Indicators of Urban Deprivation (1983), the Townsend Index (1988), the Carstairs Index (1989), the LWT Breadline Britain Index (1991) and in particular, the Index of Local Conditions (DoE, 1995) have still to be fully exploited outside of their original specific uses.

These indices of deprivation draw on the research tradition that Duncan (1969) called the 'social indicators movement' in the USA, and that developed in Britain in the 1970s. This 'movement' was largely driven by an unhappiness with the level and kinds of social information that were available to government, so that while there was a concern for social inequalities of many kinds, the main interest was not in social class per se. Indeed, within its broad church, the movement has encompassed researchers not only disinterested in class theory but also both those who were responsive or antagonistic to it.

It follows that the relationship between the social indicators approach and that used in class analysis is a complex one, not least because there is, of course, no more a single 'social indicators movement' than there is a single school of class analysis. On the one hand, in as far as we can legitimately speak of two fields of research, there is a considerable overlap of interests. Both are concerned with structures of social inequality, both have a normative orientation towards the reduction of these inequalities, and both to varying degrees use an occupational operationalization of social class in empirical research. This shared ground makes each of them potentially relevant to the other, and some writers have drawn freely on both approaches: Townsend's work is one example, while the Jarman, LWT Breadline, and Carstairs indices of deprivation incorporate occupational class, but add other variables as a way of achieving greater explanatory power.

On the other hand, there are substantial differences in orientation. Social indicator research has dealt with a multiplicity of discrete and specific social features. Its concerns are more fragmented than class analysis: at any one time, the focus may be poverty, health, housing, educational credentials, employment, crime, wealth, and so on. And within any one of these, the focus is likely to be narrower still: in health, for example, it could be standard mortality rate, or perceived health state, or number of hospitals beds, and so on. There is no single integrating idea.

The social indicators approach does not set class at the centre of explanations, nor see all social order as permeated by class processes. Carley's informative text, Social Measurement and Social Indicators (1981), for example, has no discussion of class either as an indicator or as a theoretical explanation of observed patterns of inequality. Indeed, this approach is not 'theoretical' in the sense that class analysis is essentially a theory of society. One view is that as far as social indicators research is concerned:
...there are no sociological theories about society in general on which a structure of indicators can at present be based. (Moser, 1973: p. 137).
Where the indicators school has confronted social class, a range of responses can be discerned. Hope (1978) sees quantification as a way of improving on crude class dichotomies, whereas Illsley (1990) has criticized the limitation of class as a variable, and Carr-Hill (1990) argues for its replacement by arbitrary statistical aggregates (as often used in government reports, e.g. DSS, 1992):

The only real solution to that problem is to compare the bottom 10% (or 20%) ... ranked by a criterion such as education or occupation. (1990: p. 397).
Even those more comfortable with the idea of class have tended to use it in an approximate, if practical, fashion:
...occupation is basically a pragmatic guide to that person's social position and his or her likely command over resources, and as such has its limitations. It is only an approximate indicator of family living standards or social position (Townsend et al., 1992: p. 14).

Indicators Of Deprivation

Notwithstanding some recent work that has seen employment in class terms, or in the context of power and contemporary party politics (e.g. Gallie, 1994; Byrne, 1995; Rustin, 1995), the indicators strand of research has, in the main, differentiated occupational class and the conditions of material and social deprivation into a variety of social indicators, as Table 1 shows.

Table 1: Indicators Used In The Main Deprivation Indices In Current Use
Indicator DoE
Townsend Jarman Carstairs LWT DoE, ILC
Census data
Total unemployment rate * * * * All levels
Male unemployment rate *
Overcrowded households * * * * All levels
Households lacking amenities * All levels
Not owner occupier households * *
No car households * * * All levels
Low social class (4&5 or SEG 11) * * *
Lone parent households * * *
Lone pensioner households * *
Under 5s *
Children in unsuitable accom. All levels
Children in low earning h/h All levels
Moving with previous year *
Limiting long term illness *
Born New Commonwealth * *
17 yr olds not in full time ed. ward/district
Non-census data
Standard mortality ratio district
Long term unemployment district
Income support recipients district
House contents insurance district
Low GCSE attainment district
Derelict land district

The range of variables included (and also carefully excluded: see Payne, J, 1995a) offers something more than class by itself. The indices can be used as a whole, or their components disaggregated: the Index of Local Conditions (ILC) is designed to contain within it housing and economic sub-indices. No one variable is present in the same form in all six indices, although all use a variation on unemployment and a measure of housing quality is present in all except the LWT Breadline Britain Index. The latter and the Townsend index are specifically concerned with poverty, while the Carstairs Index was developed as an alternative to using social class on its own as a means of explaining national differences in mortality rates. The Jarman Underprivileged Area Score on the other hand was constructed to measure factors affecting workloads of GPs.

Indicators are thus at once more complex and more simple, offering greater specificity, more detail, enhanced quantification and a wider range of dimensions, but bringing less ideological baggage or theoretical insight than a conventional class analysis approach. Reactions to social indicator research are therefore likely to be strongly influenced by one's initial position on class analysis: it may be seen as anything from an antagonistic paradigm to a new ally. If social indices are regarded more as a tool than an alternative perspective (the authors' present position), index-defined patterns of social inequality can be treated as straightforward descriptions of current social conditions. In practice, however, work other than that for governmental agencies often goes beyond this, incorporating conceptual frameworks even if they are not always explicitly theoretical. For example, Townsend's influential definition of relative deprivation relates the absence of resources that constitutes deprivation to the possession of resources that indicates 'non-deprivation':

People can be said to be deprived if they lack the....conditions, activities and facilities which are customary, or at least widely encouraged and approved, in the society in which they belong.(1987: pp. 125- 6)
Relative deprivation in Townsend's formulation (1987) is not very different from the view of class which emphasizes the distributional aspects of class outcomes as exemplified by Westergaard (1995). In other words we might wish to regard these indicator-defined inequalities as no more than manifestations of class; that despite any intentions by some of their creators to break free from a class framework, the indices are really just class presented in another way. If one starts from such a class analysis position the question becomes to what extent can the indices be used as supportive evidence of continuing class differentials?

In addition to the detailed exploration of deprivation patterns that this implies, it is also necessary to account for the mechanisms that presumably link inequality to class. What are the connections between market and work situations (class positions) and scores on the indices? To what extent is all social inequality to be seen as a class product?

Methodological Issues In Studying Local Patterns Of Inequality

As a step towards answering such questions, this section illustrates local patterns of inequality to demonstrate some of the potential of the deprivation index approach. Associations between variable scores for a number of 1991 Census wards are reported: it is a social area analysis, not a person-by-person analysis. It is important to remember that Census data used in the indices are not released as individual cases: to treat them as if they were would involve ecological fallacy (see Hope's note on this and the associated tendency for correlations to appear stronger than is really justified: Hope, 1978: pp. 354-8). Obviously, an analysis of census wards is a less direct method of exploring how class and deprivation are connected than one based on individual or household data. More recently, individual and household level census data have been made available for the first time as Samples of Anonymised Records (SARs). These have been extracted in two formats: a 2% sample of individual level data and a 1% sample of households (see Dale and Marsh, 1993: pp. 295-311 for full details). However, for our purposes, the SARs have three disadvantages. First, most of the data used in index construction are based on the full census count, whereas the SARs rely on 1% and 2% samples.

Second, while the SARs enable us to categorize individuals, this is at a geographical level that cannot be related to actual local circumstances (such as changes in local housing policy, employment opportunities or educational provision). Area analysis at ward (or enumeration district, or post code sector) level does offer this prospect, but with the weaker and potentially erroneous specification of how individuals experience particular deprivations.

Third, because the smallest available geographical SAR data-set is the local authority district (using the Individual 2% sample), this would not allow us to engage directly with the deprivation index tradition, which is based on the lower level units of the census geography. Deprivation indices are explicitly designed as 'indices of deprived areas': the Jarman Index, for example, is actually the Jarman Underprivileged Area score and the ILC is the Index of Local Conditions and they do not allow for individual or household- based scores in their formulae. The first step should be to explore the utility of the indices in class analysis on their own terms.

This means starting with the fact that these indices are designed to be nationally comparable and are readily available at census ward level or can be easily calculated at this level, using published national ward level means and standard deviations for the component variables (see Gordon and Forrest, 1995). While it is desirable to go to a smaller area than the ward (for reasons given in Payne, 1995a: pp. 73-4; and Payne 1996), this is not always a practical choice. It would be possible to calculate the indices for the lower level enumeration districts (EDs) but, for national comparisons to be made, this would first involve the very extensive task of calculating the means and standard deviations of each of the individual variables used for the 100,000 or so EDs. A next step might then be to construct a new deprivation scale at the individual or household level (using the SARs, for example). A number of potential lines of analysis then become possible in future research.
Our present data are drawn from the 391 wards of Devon and Cornwall, excluding the Isles of Scilly. The South West of England has been chosen for convenience, as the authors have data for it to hand, but the sub-region represents an interesting range of local conditions. It includes the most deprived ward in England (St. Peter in Plymouth, on the Index of Local Conditions) and several of the most deprived rural districts (in Cornwall). Equally it has a number of affluent areas in South Devon, but a lower than average minority ethnic group population (see Payne, J, 1995a and b). However, we are not primarily interested here in showing where deprivation is to be found in order to plan for its amelioration, nor with the levels of deprivation except in so far as these affect the interconnection of the components that we are seeking to understand. The arguments that we have developed could in principle have been built on data from any sub-region of the country.

It follows that while this is an area analysis (for the reasons given above), it is not an area analysis in the sense that we wish to identify something occurring in a particular location, or which is a product of place per se, or which requires a policy intervention on a local basis. In this respect, we differ from many of the cases discussed in Bulmer's extensive treatment of the ecological fallacy question (1986: pp. 235-45). However, we are demonstrating that the distribution of index scores, variables expressing deprivation and class composition are associated over 391 cases (wards), and we infer from this that some kind of relationship exists between them that will be manifested at individual level.

While we accept Bulmer's warning that 'the only reasonable assumption is that an ecological correlation is almost certainly not equal to its corresponding individual correlation' (1986: p. 226), we attach importance to two features of our findings. First, they establish the necessary but not sufficient prior conditions that not only do deprivation scores and class composition vary from place to place, but they have a strong tendency to vary in similar ways. This does not establish a causal relationship but it does markedly reduce the possibility that no relationship exists. If an ecological fallacy exists, it seems likely to exist in a consistent fashion across 391 research sites. Second, by using factor analysis (rather than, say, cluster analysis, which would sort wards and so emphasize the area aspect of the research), it is possible to show how variables are interconnected across the whole of the data-set.

As our aim is to demonstrate the range and potential of the deprivation indices approach, and to begin to explore how deprivation and class may be brought together in an intellectually coherent way, our inability to resort to the individual level is a price that we believe to be worth paying. Work currently in progress on enumeration districts, the OPCS Longitudinal Study and the SARs will attempt to address the individual experience of deprivation, using the framework of the inter-correlations and factors identified below. Indeed, one might wish to make the added claim that one can make little sense of class or deprivation as purely individual-level phenomena: both are systems-level social processes, within which actors act as individuals and in collectivities. While the interaction between levels remains a problem with which we must continue to wrestle, a macro-sociological approach must retain a valid place in conceptualizing the problem. We are not so rich in ideas and data that we can afford to disregard whole bodies of work, whether on class analysis, or on the area patterns of deprivation.

Thus, for the practicalities of the present analysis, we are concerned with the indices like the ILC, their component variables and with social class. Wards are given their nationally comparable scores on each of the four most relevant indices (Townsend, LWT, Jarman and ILC), their unstandardized prevalence (%) on each variable that the indices contain (an analysis using standardized scores produced very similar results), and scores for social class composition based on the economically active population. Rather than including the whole range of classes, class is operationalized using the Registrar-General's Socio-Economic Groups (SEGs); one measure is for the extent of a ward's 'middle-classness' using SEGs 1-5, 13 and 14 (managers, professionals, semi-professionals, and farmers: just over 30% in the data) and a second is for its 'working-classness' using SEG 11 (unskilled manual workers; around 6% in the data). It is recognized that there are limitations in representing class by using the top and bottom of the SEG range in this way. In addition, several alternative scores such as for types of unemployment or illness have been included. The main thrust of the analysis is an examination of the inter- correlations between sets of these scores across wards.

Readers who are unfamiliar with the indices and how the dimensions of deprivation they depict inter-correlate may like to see this.

It should be noted that this set of variables includes not only prevalence measures of the dimensions of deprivation (for example, unemployment; overcrowded housing; children in low- earning households) but also variables which identify at-risk groups (lone pensioners and lone parents, for example). Such at-risk groups have been widely identified as being likely to experience some form(s) of deprivation, but not all do. Indices which include such groups as indicators can be criticized on two grounds. First, there is a danger of both over-counting and double-counting, since all would be counted as deprived on the specific indicator and some would be counted again on other indicators. More recent indices (ILC, Townsend) specifically exclude at-risk groups for these reasons and rely on the more direct prevalence indicators, since those in an at- risk group will be counted if they actually experience deprivation.

A second criticism of the inclusion of such groups is that of circularity: by operationally defining lone pensioners, say, as a dimension of deprivation in an index, one should not be surprised to find that the resulting index shows a close association between deprived areas and areas with high proportions of lone pensioners. This is illustrated by the high correlations found between 'no car' and the Townsend index and 'children in low earning households' and the ILC (see Table B). Similarly, our analysis includes class variables which are in part already contained in the LWT and Jarman indices, but the risk of circularity here is lessened by the effect of the other variables which make up these two indices.

Urban and Rural Differences In Deprivation

A number of commentators have drawn attention to the fact that the indices work better for urban areas than for rural ones. A major factor in this is that urban wards are smaller in area and tend to be socially more homogenous than rural ones (Payne, J., 1995a). Because of the dispersed nature of populations and the wider social mix in the larger rural areas, deprivation is often disguised at ward level. Rural deprivation is distributed across spatial units.

Rural deprivation is also likely to take different forms. Studies reported by Shucksmith (1990), Room (1993) and Cloke et al. (1994) suggest that agricultural labour market restructuring and the limited range of wider employment opportunities have exacerbated the outflow of young people, leaving behind an elderly population in danger of lacking community support. A further factor is the lack of affordable homes, either because they do not exist or because houses have been bought by commuters or for holiday homes. Here again, those of working age are forced out of the area with obvious consequences for local services, social networks and community support. Thus the causes of rural deprivation, and its configurations are different, but the dimensions on which it can be measured are largely the same as for urban areas: income, unemployment, and housing.

To allow for this, wards have been allocated to one of three categories: those in urban areas, those in small towns, and those in rural areas with a population density of less than 10 persons per hectare. This latter cut off point was derived from an inspection of the data to distinguish between rural areas and small towns which do not have an agricultural labour market or an absence of local authority housing, but which would not necessarily qualify for categorisation as urban. (We do not wish to get into the debate about definitions of 'rural' here: see Payne, J., 1995b: pp. 67-8).

In the urban areas, a number of indicators show pairwise correlations of over 0.6, as do several of the class variables (in Table 2, correlations of more than 0.6, and 0.5 are underlined). There are also more correlations with higher values like 0.8 and 0.7. The indices show up as a block in the lower part of the table, correlating strongly with everything except some of the education and class variables. The strongest relationship of the class indicators with the indices is male SEG 11, and the LOCLASS variables show more clearly than do the HICLASS trio. This is also true of the relationships between class and the separate indicators of specific types of deprivation.

Table 2: Correlation Matrix For Urban Wards
CHLOE 1.0000
LONEPAR 0.8313 1.0000
CARPERH -0.8181 -0.5847 1.0000
NOTOO 0.8044 0.7760 -0.8269 1.0000
OVERCR 0.8162 0.7593 - 0.7557 0.8383 1.0000
PREMLTI 0.8120 0.6005 -0.7496 0.7516 0.6261 1.0000
SMR 0.7354 0.6258 -0.8475 0.8306 0.7605 0.7133 1.0000
EDPART 0.6077 0.6398 -0.5819 0.6484 0.6110 0.6063 0.4814 1.0000
QUALS -0.5204 -0.5454 0.2838 - 0.2261 -0.2602 -0.4332 -0.1636 -0.5683 1.0000
UNEMPL 0.6970 0.4274 -0.6956 0.5325 0.6395 0.5291 0.5461 0.2868 - 0.3237 1.0000
LOCLASS 0.8434 0.8369 -0.7071 0.8059 0.7834 0.7603 0.6931 0.8018 - 0.5277 0.5053 1.0000
LOCLASSM 0.8310 0.7413 -0.7953 0.8761 0.7925 0.7471 0.7485 0.7469 -0.4487 0.5809 1.0000
LOCLASSF 0.7228 0.7843 -0.5256 0.6155 0.6478 0.6569 0.5344 0.7258 -0.5223 0.3650 1.0000
HICLASS - 0.6188 -0.7000 0.3981 -0.4198 -0.4426 -0.4837 -0.2901 - 0.6985 0.9034 -0.3351 1.0000
HICLASSM -0.7012 -0.7502 0.5194 - 0.5291 -0.5157 -0.5547 -0.4236 -0.7324 0.8709 -0.4003 1.0000
HICLASSF -0.4250 -0.5409 0.1753 - 0.2018 -0.2775 -0.3229 -0.0577 -0.5617 0.8523 -0.2032 1.0000
TOWNSEND 0.8943 0.7175 - 0.9279 0.9204 0.9004 0.7717 0.8306 0.4814 -0.2863 0.7607 0.7857 0.8656 0.5943 - 0.4266 -0.5352 -0.2135
LWT 0.8989 0.7681 - 0.9231 0.8643 0.8643 0.8385 0.8668 0.6789 -0.3566 0.6632 0.8433 0.8954 0.6661 - 0.5096 -0.6153 -0.2912
ILC 0.8239 0.5654 - 0.9164 0.7678 0.7678 0.6647 0.7525 0.4947 -0.2436 0.7737 0.6286 0.7464 0.4329 - 0.3374 -0.4568 -0.1208
JARMAN 0.8761 0.6646 - 0.9618 0.8159 0.8159 0.7554 0.8674 0.5671 -0.2939 0.7370 0.7356 0.8028 0.5651 - 0.3827 -0.5009 -0.1642

Children in low-earning households (CHLOE - for definitions see Annex) is the most distinctive indicator, having strong correlations with all of the other important indicators except females in the upper SEGs (HICLASSF). This is now followed by 'not owner-occupier' (NOTOO).

When we carry out the same exercise for small towns, the most obvious difference is the lower level of some of the correlations. Only just over half of the possible cells exceed r=0.6, in contrast to 80% in the previous table. The education, unemployment and health variables that showed up clearly in the urban areas become less visible. Among the class variables, a middle- class negative relationship with deprivation is the stronger effect, but the overall LOCLASS indicator also retains high correlations with the separate variables. The housing indicators remain, and are joined by over-crowding; long-term illness is also associated with LOCLASS. Both female class scores are again lower than the males, but there is less gender difference between the lower-class variables.

Full Correlation Matrix

CHLOE and LONEPAR have most high correlations among the specific deprivation indicators. The housing variables also show up strongly, while illness is correlated with the indices, income and non-owner- occupation. The income indicator is linked with housing, illness and the 4 indices. Unemployment appears only associated with the ILC and Townsend (which contain it as a component in their construction), or through CHLOE and LONEPAR.

Turning now to the most rural wards, the lack of underlined values (less than 20%) in Table 3 tells its own story. Deprivation is less visible in the larger, more heterogeneous, countryside wards, and its inter- correlations are much lower. The lower class has a weaker association with deprivation than the middle class has with its absence, but even here there are at best only two correlations over 0.5 for the male middle-class indicator; this is only one-quarter of the number of high correlations in the small towns and even fewer than the urban areas where the lower class is more important. HICLASS is strongly negatively correlated with the income proxy and qualifications, as is the male version of this. HICLASSF is linked to qualifications. No other large class/deprivation correlations were found. Among the indices, Jarman above 0.6, and Townsend at over 0.5, are the only two to show strong correlations.

Table 3: Correlation Matrix For Rural Wards
CHLOE 1.0000
LONEPAR 0.7178 1.0000
CARPERH -0.6411 -0.5149 1.0000
NOTOO 0.2908 0.4122 -0.2952 1.0000
OVERCR 0.4365 0.4409 - 0.3130 0.3079 1.0000
PREMLTI 0.5652 0.3257 -0.4179 0.0708 0.1991 1.0000
SMR 0.2481 0.2342 -0.3249 0.1053 0.2589 0.3403 1.0000
EDPART 0.2783 0.3070 -0.2534 0.1779 0.1806 0.2560 0.2678 1.0000
QUALS -0.4190 -0.3902 0.5138 - 0.2096 -0.2884 -0.3486 -0.3083 -0.4797 1.0000
UNEMPL 0.4196 0.2173 -0.3565 0.0420 0.2764 0.3155 0.0666 0.1025 - 0.2381 1.0000
LOCLASS 0.3121 0.2823 -0.4033 0.2355 0.1657 0.2547 0.1733 0.1660 - 0.3665 0.1983 1.0000
LOCLASSM 0.3167 0.2736 -0.3782 0.1493 0.1986 0.2552 0.1001 0.1215 -0.2841 0.2556 1.0000
LOCLASSF 0.1805 0.1602 -0.2409 0.2022 0.0559 0.1564 0.1519 0.1513 -0.2912 0.0609 1.0000
HICLASS -0.4571 -0.4526 0.6332 -0.2134 - 0.3036 -0.3090 -0.3435 -0.3992 0.7148 -0.1845 1.0000
HICLASSM -0.4752 -0.4817 0.6434 -0.2251 - 0.3355 -0.2804 -0.2959 -0.3696 0.6428 -0.2266 1.0000
HICLASSF -0.2768 -0.2468 0.3826 - 0.1223 -0.1552 -0.2652 -0.2972 -0.3317 0.6130 -0.0627 1.0000
TOWNSEND 0.7486 0.6358 - 0.7066 0.6088 0.7002 0.4776 0.3210 0.2901 -0.4835 0.4794 0.3803 0.3701 0.2212 - 0.5202 -0.5429 -0.3121
LWT 0.5642 0.5246 - 0.6611 0.7324 0.3975 0.3822 0.2756 0.2487 -0.4272 0.2647 0.3699 0.2896 0.2776 - 0.4572 -0.4692 -0.2834
ILC 0.7013 0.4462 - 0.5312 0.3133 0.5186 0.5910 0.2976 0.3104 -0.3980 0.5049 0.2623 0.2920 0.1337 - 0.3519 -0.3680 -0.2216
JARMAN 0.6817 0.6544 - 0.8158 0.3848 0.4639 0.3831 0.2900 0.2162 -0.4305 0.3815 0.5176 0.5045 0.2953 - 0.5921 -0.6063 -0.3521

One also looks in vain for many relationships between other individual variables. Apart from the Townsend and Jarman associations, CHLOE is correlated with single-parent households, income and illness, the same combination as for LONEPAR. Most of the other variables fall well short of any 0.5 correlation. We must conclude that the combination of distinctive labour market and housing circumstances referred to above, plus the technical limitations of area-based approaches that fail to pick up cases of deprivation, restrict what can be said about rural deprivation.

To summarize, the variation found between the three types of areas justifies the urban/rural distinction: the three specific patterns would have remained hidden if the analysis had operated only at the county- wide level. Worse, the numerical preponderance of rural wards would have out- weighed the real distribution of population between rural areas and more urban types of location. While there are doubts on technical grounds about what can be discovered in rural areas, the size and population of the urban wards and those in the small towns are very similar. Any differences between the latter two are unlikely to be artefactual. As most people in England live in these more urban areas, the patterns reported here are likely to be more typical of the national picture.

Factors in Deprivation

While an investigation of how pairs of variables correlate is an important part of understanding deprivation patterns, it is also necessary to examine how combinations of variables are inter-linked. One technique for doing this is factor analysis using the principal component method (PCA). This is a widely used exploratory technique that has been successfully used in the analysis of large data-sets, including census data (Moser and Scott, 1961; Holterman, 1975; DoE, 1995), and has been identified as a suitable analytical method for deprivation studies.

Factor analysis is a very powerful tool for revealing underlying structures in dataÉand can throw light on complex concepts such as deprivation. (Folwell, 1993: p. 38)
PCA has been described as 'arguably one of the most useful ways of exploring a covariance structure' (Taylor, 1977: p. 101) and can be usefully employed in 'the analysis of a set of data where the structure is not explicitly causal, but exists merely in that there is a stable pattern of covariance' (1977: p. 90). The aim of PCA is to simplify data into more easily understood components or factors and it is particularly appropriate where there are a large number of inter-correlated variables with complex relationships between each other, as here. Further, unlike regression models, PCA does not impose a predetermined causal model upon the data: the final outcome depends entirely upon the relationships between the variables. When a group of variables has a great deal in common, an underlying factor may be said to exist. Thus, as found in the correlation matrices of the individual deprivation indicators reported above, the inter-correlations may indicate the possibility of one or more factors, which might simplify and clarify the conceptual model. Factors may be selected to be independent of each other, by using an orthogonal method, or related, by using oblique rotation.

The resulting factor scores can then be used to group cases using cluster analysis. Although both techniques are classificatory methods, it is important to distinguish between their approaches: factor analysis is used primarily to group variables, while cluster analysis is used to group cases on the basis of selected variables. Even though cluster analysis is often used as an alternative exploratory method to factor analysis, it is more common in large and complex data sets to first 'reduce' or simplify the data using factor analysis (thus reducing the number of variables) before clustering, so that the results are more interpretable i.e. what have these cases got in common? Cluster analysis:

...should be used in association with other multivariate analysis methods such as principal componentsÉIn this way greater insight into the structure of the data will be obtained. The practice of simply applying one particular technique to a set of data, and accepting the solution produced is not to be recommended. Clusters should be evaluated in the light of the results of other techniques. (Everitt, 1977: p. 86)
This use of cluster analysis as a second stage multivariate analysis is also referred to by the same author in a later discussion of the method (Everitt, 1993).

Our main reason for using PCA was thus that it appeared to be the most appropriate method in this first stage of multivariate analysis, because we were interested in the relationships between variables rather than between wards. However, the use of cluster analysis had been suggested at an early stage in the research. When this was undertaken (selecting 2, 3 and 4 cluster solutions), it produced less valuable information about the relationships between the variables than that from the bivariate correlations (i.e. the mean values of the variables for the most deprived group of wards were as expected). All of the variables except for the indices were included in the PCA model (i.e. not only those with high correlations to another variable, of the kind discussed up to this point). As we shall see, this sometimes creates factors that make statistical sense, but are hard to explain in sociological terms.

The results for each of the three areas are shown in Tables 4, 5 and 6: factor loadings of 0.5 or greater are underlined as these are used in our interpretation of the meaning of the factors (although some writers suggest a less cautious approach: Folwell, 1993: p. 36).

Table 4: Urban Factor Loading Matrix
Factor 1 Factor 2 Factor 3 Factor 4
LOCLASS 0.83 0.22 -0.40 - 0.13
LOCLASSM 0.84 0.33 - 0.23 -0.10
LOCLASSF 0.69 0.09 -0.49 - 0.12
HICLASS - 0.41 -0.12 0.84 0.18
HICLASSM -0.52 -0.16 0.75 0.15
HICLASSF -0.19 - 0.05 0.88 0.20
CAR2+ -0.76 -0.51 0.07 - 0.21
CARPERH - 0.77 -0.54 0.02 -0.24
CHLOE 0.74 0.48 - 0.35 0.03
CHUNSACC 0.55 0.54 0.09 0.13
EDPART 0.71 0.05 - 0.43 -0.14
LACKAMEN 0.14 0.51 0.73 - 0.09
LAHOUSE 0.80 -0.12 - 0.45 -0.04
LONEPAR 0.73 0.18 -0.43 - 0.31
LONEPEN - 0.00 0.12 0.14 0.96
NOCAR 0.76 0.56 0.01 0.26
NOTOO 0.92 0.28 -0.01 - 0.07
OVERCR 0.75 0.47 - 0.07 -0.27
PREMLTI 0.75 0.25 -0.31 0.36
QUALS -0.21 -0.15 0.90 0.02
SMR 0.84 0.31 0.07 0.17
UNEMPL 0.28 0.92 - 0.16 0.03
UNTOTF 0.04 0.88 0.03 0.03
UNTOTM 0.36 0.84 - 0.20 0.03
YNGGTS -0.02 0.10 -0.75 0.14
YNGUNEMP 0.41 0.78 - 0.13 0.01
Eigenvalue 14.09 4.61 1.89 1.42
% Variance explained 54.20 17.80 7.30 5.40
Cumulative variance 54.20 72.00 79.30 84.70

In the urban areas, a four- factor solution accounts for nearly 85% of the variance. The first factor is a large combination of the variables already identified as having high correlations between pairs (e.g. LONEPAR, CHLOE, low car and house ownership, illness) with low social class. This explains more than half the variance. The second factor is a combination of 4 unemployment variables with low income and lack of housing amenities, while the fourth factor is lone pensioner households. The third factor is a little more complicated: it combines the middle-class and educational indicators with lack of housing amenities. A possible explanation is that this identifies city areas with mixes of housing stock in which 'down-town intellectuals' live close to other people having different life- styles.

The pattern of factors for small towns shown in Table 5 is quite different. 80% of the variance is explained by six factors, the main one being a combination of negative loadings on the middle class and education variables and positive loadings on lone parents and SMR. That they appear in this way suggests that the class ecology of urban areas and small towns is different.

Table 5: Small-Town Factor Loading Matrix
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6
LOCLASS 0.20 0.42 0.12 0.12 0.82 -0.01
LOCLASSM 0.13 0.16 0.27 0.17 0.75 0.04
LOCLASSF 0.22 0.57 - 0.09 0.05 0.62 -0.03
HICLASS -0.91 -0.23 - 0.08 -0.08 -0.20 -0.00
HICLASSM -0.84 -0.26 0.25 -0.15 - 0.16 0.10
HICLASSF -0.83 - 0.13 0.03 0.05 -0.22 0.03
CAR2+ -0.39 - 0.30 -0.66 -0.19 -0.18 -0.34
CARPERH -0.28 - 0.37 -0.71 -0.20 -0.20 -0.37
CHLOE 0.43 0.67 0.23 0.43 0.23 0.02
CHUNSACC 0.08 -0.03 0.80 0.30 - 0.13 -0.05
EDPART 0.62 0.24 0.14 - 0.03 0.06 0.41
LACKAMEN -0.01 -0.09 0.15 0.27 0.06 0.84
LAHOUSE 0.34 0.77 0.10 -0.03 0.27 -0.21
LONEPAR 0.50 0.71 -0.10 0.10 0.22 - 0.02
LONEPEN - 0.24 -0.22 0.87 -0.04 -0.01 - 0.06
NOCAR 0.19 0.40 0.72 0.21 0.12 0.38
NOTOO 0.30 0.73 0.43 0.03 0.24 0.08
OVERCR 0.29 0.71 0.02 0.25 0.19 0.34
PREMLTI 0.31 0.26 0.60 0.30 0.37 0.06
QUALS -0.81 -0.25 - 0.12 -0.11 -0.17 -0.03
SMR 0.53 0.33 0.43 0.03 - 0.16 -0.02
UNEMPL 0.12 0.01 0.22 0.94 0.10 0.08
UNTOTF -0.06 0.20 0.03 0.76 - 0.18 0.03
UNTOTM 0.21 - 0.10 0.25 0.79 0.23 0.09
YNGGTS 0.31 0.08 -0.21 - 0.06 0.56 0.26
YNGUNEMP -0.06 0.29 0.11 0.71 0.24 0.31
Eigenvalue 10.86 3.81 2.15 1.69 1.27 1.06
% Variance explained 41.80 14.60 8.30 6.50 4.90 4.10
Cumulative variance 41.80 56.40 64.70 71.20 76.10 80.20

The second factor partly resembles the urban first factor, consisting of CHLOE, LONEPAR and the housing variables, whereas the low-income surrogates show more clearly in a third factor with lone pensioners and 'children in unsuitable accommodation', while low class (apart from females) is a separate fifth factor. Unemployment constitutes the fourth factor, while poor housing amenities comprise the other. This pattern may suggest that in some small towns there are wards containing poor people, the sick, and the elderly living on their own, in private rented accommodation, and other wards with deprived households who are associated with over-crowding and living in local authority housing (a line of analysis now being followed up by the authors). Unemployment, poor housing, and low class are also factors in their own right, whereas in cities only unemployment (and lone pensioners) seems to have an explanatory power separate from the cluster of multiple deprivations identified above as the first factor in the urban areas

Unemployment, low class, and the poverty/single elderly factor are also identified in the rural areas shown in Table 6, but here it takes seven factors to reach 72.8% of explained variance.

The main factor is again HICLASS and education. The remaining three factors are lone-parent and unemployed parent households living in public housing; lack of housing amenities and poor health; and unemployed females. It is not clear why the latter is separate: female unemployment may be a distinctive feature of rural labour markets which lack office work and also depend on seasonal employment in tourism, whereas male and youth unemployment are found in all three types of area.

Table 6: Rural Factor Loading Matrix
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7
LOCLASS -0.24 0.13 0.14 0.14 0.92 0.03 0.01
LOCLASSM -0.15 0.09 0.15 0.45 0.63 0.06 - 0.19
LOCLASSF - 0.22 0.10 0.07 -0.19 0.80 0.11 0.18
HICLASS 0.81 - 0.25 -0.21 -0.11 -0.34 -0.00 0.08
HICLASSM 0.71 -0.26 - 0.27 -0.22 -0.26 0.07 0.10
HICLASSF 0.71 - 0.12 -0.02 0.09 -0.35 -0.13 0.01
CAR2+ 0.44 -0.75 - 0.27 -0.26 -0.12 -0.02 -0.00
CARPERH 0.41 - 0.78 -0.32 -0.26 -0.11 -0.01 0.00
CHLOE -0.27 0.32 0.50 0.52 0.06 0.23 - 0.08
CHUNSACC 0.13 0.63 - 0.00 0.17 0.07 0.20 0.31
EDPART -0.66 - 0.10 0.22 -0.05 -0.08 0.25 0.13
LACKAMEN 0.02 -0.15 0.05 0.21 0.05 0.72 0.10
LAHOUSE -0.28 0.18 0.81 -0.03 0.13 -0.13 - 0.07
LONEPAR - 0.32 0.10 0.70 0.30 0.03 0.09 -0.08
LONEPEN 0.02 0.89 -0.08 - 0.01 0.10 -0.40 0.01
NOCAR -0.33 0.78 0.38 0.24 0.10 0.02 0.00
NOTOO 0.01 0.08 0.83 -0.16 0.17 0.05 0.19
OVERCR -0.13 -0.00 0.53 0.30 0.01 0.29 0.06
PREMLTI -0.22 0.27 0.07 0.39 0.10 0.53 - 0.15
QUALS 0.79 -0.13 - 0.16 -0.12 -0.14 -0.15 -0.08
SMR -0.28 0.26 0.14 - 0.10 0.04 0.59 -0.12
UNEMPL -0.11 0.15 0.01 0.75 0.05 0.06 0.57
UNTOTF -0.05 0.10 0.07 0.21 0.04 -0.00 0.83
UNTOTM -0.12 0.15 - 0.03 0.82 0.03 0.08 0.22
YNGGTS -0.37 - 0.02 -0.10 0.20 -0.04 0.42 0.21
YNGUNEMP 0.09 0.17 0.16 0.63 0.03 0.25 0.05
Eigenvalue 8.59 2.75 2.08 1.80 1.50 1.14 1.06
% Variance explained 33.00 10.60 8.00 6.90 5.80 4.40 4.10
Cumulative variance 33.00 43.60 51.60 58.60 64.30 68.70 72.80

This factor analysis points to four main conclusions about deprivation. In all areas, both class and unemployment are identifiable factors. An explanation of deprivation must therefore take them into account. Second, poverty takes different configurations, highly inter-connected in cities and more disaggregated elsewhere; we need to consider why this should be. Third, households with small children where the single, or both, parents are not in full-time paid employment are a specific locus of poverty. Fourth, pensioners living on their own constitute a separate category, with a more complex association with deprivation that requires further consideration.

Reconsidering Class and Deprivation

The central objective of the article has been to promote dialogue between class analysis and deprivation research. Current low levels of dialogue have been attributed to two sources. On the one hand, class analysis in several formulations has yet to produce a coherent, convincing or empirically fruitful heuristic device for explaining deprivation. Most class analysis has concentrated on production relations or global systems, rather than the conditions and causes of poverty and social exclusion. On the other hand, deprivation research sits in a social policy/social indicators tradition which may be technically sophisticated but has been theoretically limited. It is therefore perhaps not to be unexpected that a lack of dialogue between the two approaches has been the result.

We therefore began by suggesting some of the limitations and the potential of conventional class analysis, before moving on to report some of the richness of recent social indicator research, with the intention of encouraging readers in one or other camp to reflect on what the other camp might have to offer. Although area analysis is fraught with technical problems, we attempted to pick our way through these in order to present contemporary sub-regional data on patterns of deprivation. To the extent that we have been successful, we can now return to the initial problem, and reflect on whether such patterns can be said to be part of social class, caused by the class system, help us to understand class or, in turn, whether deprivation patterns extend our understanding of social class.

It may be helpful to restate some of the main patterns. Allowing for some simplification, there are substantial differences between urban, small town and rural areas:

Apart from telling us something about deprivation per se, do these results tell us anything about class? In the first place, we can reject any idea that deprivation and poverty are just outward manifestations of class. While it is true that the middle classes are conspicuous by their absence from deprived areas, the several dimensions of deprivation - poor housing, poverty, ill health, household composition - do not automatically follow the lower classes. Indeed, the dimensions of deprivation do not all behave in the same way, so one may conclude that, even if there is a class connection, certain aspects of deprivation operate at least in part independently of class.

Among these operations a life-cycle effect is visible. The association of poverty with households containing children (and unemployment) and lone pensioners suggest a more prosaic biological component, even if part of the explanation lies in cultural norms about childhood, 'family' wage rates, gender, and demand for labour. Although the factor analysis was not set up to produce an age or gender effect, these are clear in the results. On the other hand, it would be wrong to represent these findings as excluding an association between deprivation and class. The advantaged position of the educated middle classes and the disadvantaged position of the unskilled workers - both as a separate factor and linked in urban areas to other indicators of deprivation - is evident.

Turning therefore to more specific components of class theory, we earlier showed some sympathy in our introductory discussion with the general orientation of the consumption sector approach. However, the deprivation findings do not fit comfortably into a residualization model. We find complex but separate poverty groupings, rather than neat coherent ones in the residual category of local authority housing, across all three ward types. In addition we find class both correlating with other variables and operating as powerful independent factors. While consumption theory may offer a political explanation that should not be lost, the salience of class needs to be re-established within its framework.

The key to this lies in seeing how both income level and risk of unemployment run through so much of the other findings. As we argued above, to the extent that income, unemployment and occupationally defined class are ultimately based in production (i.e. market) relations and if they are associated with risk of deprivation, then market relations underpin deprivation patterns. In this way, although class as a direct basis for collective action may lose its significance (if indeed it ever did possess it!) it can be shown to affect the new alternative 'consumption' forms of collectivity by assisting in the creation of specific types and distributions of deprivation.

It follows that we need to recognize the primacy of class position in the sense of a capacity to exchange labour (or property) in the market, for income, in classic Weberian mode. This capacity includes both the kinds of labour that can be offered for sale, and the demand for it which creates employment or unemployment. This does not mean that we equate unemployment with unskilled labour, but rather that occupation, industry (and employment status) affect propensity to become unemployed, as Gallie and Vogler (1993) argue. Without the demand opportunity for employment there can be no exchange of labour for income, while type of labour also carries with it different levels of income.

By primacy we do not mean exclusivity: other factors like life-cycle in the form of membership of households containing children, or being a lone pensioner are central to deprivation when they interact with current or former class position. Similarly, access to local authority housing is dependent both on the state's supply, and on the locations of residence and employment. Indeed, employment itself is also spatially differentiated; the complexity of the rural patterns of deprivation is not simply artefactual. Thus despite the logical primacy of class, we cannot regard deprivation as just an adjunct or by-product of class.

To go one step further, our findings that several indicators of deprivation are associated confirm that there is something distinctive about it. Deprivation has a class-connected but independent existence, because it is a way of life. Its potential as a source of group identity may be weak because of the way poverty incapacitates, but its existence as a marker for a stigmatized out-group plays a prominent role in media and political debate (as recent remarks about 'squeegee merchants' testify). It is this essence and appearance of distinctiveness that has given credence to recent theories of the underclass, which posit the existence of a self-contained group with its own values dependent on, but living outside of, the mainstream of society. The term 'underclass' is useful to the extent that it denotes 'status exclusion' (Morris, 1994), or exclusion from the rights of 'social citizenship' (Lister, 1991; Crompton, 1993), but as a concept, its explanatory potential is limited. Deprivation is distinct, but it results from a lack of opportunities as already explained. It is difficult to see how the complex patterns we have discovered can entirely be reduced to 'inadequate parenting', 'fecklessness' and 'voluntary unemployment', as suggested by the New Right (Murray, 1990, 1994; Marsland, 1994) or indeed to the depravity of Marx's scum, offal and refuse of society. The more sophisticated, market- based model that we have begun to sketch has much more to offer both in its own right, and in where it may lead. For example, the social exclusion entailed in being deprived cannot but be relevant to differential life-chances and mainstream class analysis topics such as social mobility, access to education, and quality of life as manifested by poor health.

The deprivation patterns that we have reported should not come as a surprise in view of the massive economic and political restructuring that has taken place over the last two decades. Declining investment and deflationary economic policies have resulted in de-industrialization and rising unemployment: large sectors of the labour market have contracted. A permanent feature of society is mass unemployment affecting a third of households, with a crucial division between those in and those out of work. Our use of 'no car' as an income surrogate is doubly appropriate because this and other aspects of household consumption are intrinsic to 'relative deprivation', providing foundations for new social cleavages reflecting consumption. It is not just 'Who Gets What?' (Westergaard, 1995) but also 'Who doesn't get what?'.

Paradoxically, although we would argue that deprivation needs to be seen in its own light, it is labour-market class position itself that helps to create the alternative cleavages. The study of deprivation as a source of stratification in its own right, as an addition to more conventional class analysis, helps to account for key changes in contemporary society. This means that sociologists will need to pay greater attention to social indicators, and in particular that indices of deprivation are in future likely to assume a greater importance on the research agenda.


Some data used in this paper are taken from the 1991 Population Census., Crown copyright. ESRC purchase.
The authors are grateful to members of the Cambridge Stratification Seminar, and Alison Green and Paul Iganski, who made helpful suggestions on an earlier version of this paper and to the University of Plymouth Vice-Chancellor's Fund for financial assistance in carrying out part of the research reported.


AMIN, A. (editor) (1993) Post- Fordism: A Reader. Oxford: Blackwell.

BECK, U. (1992) Risk Society: Towards A New Modernity. London: Sage.

BROWN, P. and CROMPTON, R. (1994) Economic Restructuring And Social Exclusion. London: UCL Press.

BROWN, P. and SCASE, R. (1991) Poor Work: Disadvantage And The Division Of Labour. Buckingham: Open University Press.

BULMER, M. (1986) Social Science And Social Policy. London: Allen and Unwin.

BURROWS, R. and LOADER, B. (editors) (1994) Towards A Post-Fordist Welfare State?. London: Routledge.

BYRNE, D. (1995) 'Community And The Dispossessed Working Class', paper presented to Ideas of Community Conference, University of the West of England, Bristol, September.

CARLEY, M. (1981) Social Measurement And Social Indicators. London: Allen and Unwin.

CARR-HILL, R. (1990) 'The Measurement Of Inequalities In Health', Social Science and Medicine, vol. 31, no. 3, pp. 393-404.

CARSTAIRS, V. and MORRIS, R. (1989) 'Deprivation: Explaining Differences In Mortality Between Scotland And England And Wales', British Medical Journal, 299, 886-889.

CLOKE, P., MILBOURNE, P. and THOMAS, C. (1994) Lifestyles In Rural England. London: RDC/HMSO.CROMPTON, P. (1993) Class And Stratification: A Guide To Current Debates. Cambridge: Polity.

DALE, A. and MARSH, C. (editors) (1993) The 1991 Census User's Guide. London: HMSO.

DoE (1983) Urban Deprivation: Information Note 2. London: DoE.

DoE (1995) 1991 Deprivation Index: A Review Of Approaches And A Matrix Of Results. London: HMSO.

DSS (1992) Households Below Average Income. London: HMSO.

DUNCAN, O. D. (1969) Towards Social Reporting: Next Steps. New York: Russell Sage Foundation.

DUNLEAVY, P. (1980) Urban Political Analysis. London: Macmillan.

EDER, K. (1993) The New Politics Of Class: Social Movements And Cultural Dynamics In Advanced Societies. London: Sage.

EUROPEAN COMMISSION (1995) Trends: Employment in Europe. Brussels: SYSDEM.

EVERITT, B. S. (1977) 'Cluster Analysis' in C. A. O'Muircheataigh and C. Payne (editors) Exploring Data Structures. London: John Wiley.

EVERITT, B. S. (1993) Cluster Analysis. London: Edward Arnold.

FOLWELL, K. (1993) 'Seeking A Measure Of Deprivation - Factor Analysis And Cluster Analysis' in S. Simpson (editor) Census Indicators of Local Poverty and Deprivation. London: LARIA.

FRAYMAN, H., MACK, J., LANSLEY, S., GORDON, D. and HILLS, J.(1991) Breadline Britain In The 1990s: The Findings Of The Television Series. London: Domino Films and LWT.

GAFFIKIN, F. and MORRISSEY, M. (1992) The New Unemployed. London: Zed.

GALLIE, D. (1994) 'Are The Unemployed An Underclass?', Sociology, vol. 28, no. 3, pp. 737-57.

GALLIE, D. and VOGLER, C. (1993) 'Labour Market Deprivation, Welfare And Collectivism' in D. Gallie, C. Marsh, and C. Vogler (editors) Social Change And The Experience Of Unemployment. Oxford: Oxford University Press.

GEORGE, V. and HOWARDS, I. (1991) Poverty Amidst Affluence. London: Edward Elgar.

GOLDTHORPE, J. and MARSHALL, G. (1992) 'The Promising Future Of Class Analysis: A Response To The Recent Critiques', Sociology, vol. 28, no. 3, pp. 381-400.

GORDON, D. and FORREST, R. (1995) People and Places 2. Bristol: School for Advanced Urban Studies.

GREEN, A. (1994) The Geography Of Poverty And Wealth. Warwick: IER.

HOLTERMAN, S. (1975) 'Areas Of Urban Deprivation In Great Britain; An Analysis Of 1971 Census Data', Social Trends 6. London: HMSO.

HOPE, K. (1978) 'Indicators Of The State Of Society', in M. Bulmer (editor) Social Policy Research. London: Macmillan.

ILLSLEY, R. (1990) 'Comparative Review Of Sources, Methodology And Knowledge', Social Science and Medicine, vol. 31, no. 3, pp. 229-36.

JARMAN (1983) 'Identification Of Underprivileged Areas', British Medical Journal, vol. 286, pp. 1705-1709.

JENCKS, C. and PETERSON, P. E.(1991) The Urban Underclass. Washington DC: The Brookings Institute.

JESSOP, B. (1993) 'The Transition To Post-Fordism And The Schumpeterian Workfare State' in R. Burrows and B. Loader (editors.) Towards A Post-Fordist Welfare State?. London: Routledge.

JOSEPH ROWNTREE FOUNDATION (1995) Inquiry into Income and Wealth. York: Joseph Rowntree Foundation.

Le GRAND, J. (1982) The Strategy of Equality. London: Unwin Hyman.

LIPIETZ, A. (1992) Towards A New Economic Order. Cambridge: Polity.

LISTER, R. (1991) The Exclusive Society: Citizenship And The Poor. London: CPAG.

MANN, K. (1991) The Making Of An English Underclass. Buckingham: Open University Press.

MARSHALL, T. H. and BOTTOMORE, T.(1992) Citizenship And Social Class. London: Pluto.

MARSLAND, D. (1994) 'The Use And Abuse Of State Welfare', The Salisbury Review, vol. 12, no. 4, pp. 14-17.

MARX, K. (1962) 'The Class Struggles In France 1848- 1850,' in Karl Marx and Frederick Engels, Selected Works (Vol. 1). London: Lawrence and Wishart.

MORRIS, L. (1994) Dangerous Classes: The Underclass And Social Citizenship. London: Routledge.

MOSER, C. A. (1973) 'Social Indicators: Systems, Methods And Problems', in M. Bulmer (editor) Social Policy Research. London: Macmillan.

MOSER, C. A. and SCOTT, W. (1961) British Towns: A Statistical Study Of Economic And Social Differences. Edinburgh: Oliver and Boyd.

MURRAY, C. (1990) The Emerging British Underclass. London: IEA.

MURRAY, C. (1994) Underclass: The Crisis Deepens. London: Institute of Economic Affairs.

NOVAK, T. (1988) Poverty And The State. Buckingham: Open University Press.

PAHL, R. (1984) Divisions Of Labour. Oxford: Blackwell.

PAHL, R. (1988) 'Some Remarks On Informal Work, Social Polarisation And Social Structure', International Journal of Urban and Regional Research, vol. 12, pp. 247-267.

PAYNE, G. (1987) Mobility And Change In Modern Society. London: Macmillan.

PAYNE, G. and FORD, G.,(1983) 'Inequality And The Occupational Structure' in G. Brown, and R. Cook (editors) Scotland: The Real Divide. Edinburgh: Mainstream.

PAYNE, J. (1995a) Interpreting The Index Of Local Conditions: Relative Deprivation In Devon And Cornwall. Plymouth: Plymouth Business School, University of Plymouth.

PAYNE, J. (1995b) Regional Profile Of The South West With Special Reference To Disadvantage And Poverty. Plymouth: Community Research Centre, University of Plymouth.

PAYNE, J. (1996) Health And Deprivation In Cornwall. Plymouth: Faculty of Human Sciences, University of Plymouth, (in preparation).

POWER, A. and TUNSTALL, R. (1995) Swimming Against The Tide. York: Joseph Rowntree Foundation.

ROOM, G. (editor) (1993) Anti-Poverty Action Research in Europe. Bristol: School for Advanced Urban Studies.

RUSTIN, M. (1995) 'Community, Market And State' paper presented to Ideas of Community Conference, University of the West of England, Bristol, September.

SAUNDERS, P. (1981) 'Beyond Housing Classes' International Journal of Urban and Regional Research, vol. 8, pp. 202-27.

SHUCKSMITH, M. (1990) The Definition Of Rural Areas And Rural Deprivation. Edinburgh: Scottish Homes.

TAYLOR, C. C. (1977) 'Principal Component And Factor Analysis' in C. A. O'Muircheataigh and C. Payne (editors) Exploring Data Structures. London: John Wiley.

TITMUSS, R. M. (1958) Essays On The Welfare State. London: Unwin University Books.

TOWNSEND, P. (1987) 'Deprivation', Journal of Social Policy, vol. 16, pp. 125-46.

TOWNSEND, P. (1993a) The International Analysis Of Poverty. London: Harvester Wheatsheaf.

TOWNSEND, P. (1993b) 'Underclass And Overclass: The Widening Gulf Between Social Classes In The 1980s' in G. Payne, and M. Cross (editors) Sociology in Action. Basingstoke: Macmillan.

TOWNSEND, P., DAVIDSON, N. and WHITEHEAD, M. (editors)(1992) Inequalities In Health. London: Penguin.

TOWNSEND, P., PHILLIMORE, P. and BEATTIE, A.(1988) Health And Deprivation: Inequalities In The North. London: Croom Helm.

WESTERGAARD, J. (1992) 'About And Beyond The "Underclass": Some Notes On Influences Of Social Climate On British Sociology Today', Sociology, vol. 26, pp. 575-87.

WESTERGAARD, J. (1995) Who Gets What?. Cambridge: Polity Press.

WILSON, J. (1987) The Truly Disadvantaged. Chicago: University of Chicago Press.

WRIGHT, E. O. (1995) Interrogating Inequality. London: Verso.


Units of Analysis:

391 Census Wards in Devon and Cornwall (excluding Scillies)

Variables used in Analysis:

Class variables:
(Economically active)

HICLASS SEG 1, 2, 3, 4, 5, 13, 14
HICLASSF Female SEG 1-5, 13,14
HICLASSM Male SEG 1-5, 13,14

Deprivation IndicatorsIndices:
(See Table 1 for variables used in individual indices)

JARMAN includes Class
LWT includes Class

Individual deprivation indicators:
(Percentages unless otherwise stated)

CAR2+ 2 or more car households
CARPERH cars per household
CHLOE children in low earning households (neither parents in work or, if lone parent, not working or working part time)
CHUNSACC children in unsuitable accommodation
EDPART educational participation - 17yr olds not in FT education
LACKAMEN lack of basic amenities
LAHOUSE local authority housing
LONEPAR lone parent households
LONEPEN lone pensioner households
NOCAR no car households
NOTOO not owner occupier
OVERCR overcrowded households(over 1 persons/room)
PREMLTI 16-pension age - long term sick
QUALS Qualifications - diploma, degree +
SMR premature death rate (16-pension age - SMR)
UNEMPL unemployed - total
UNTOTF female unemployed
UNTOTM male unemployed
YNGGTS 16-24 ec active - on government training scheme
YNGUNEMP 16-24 ec active - unemployed

Copyright Sociological Research Online, 1996