Research Policy and Planning: The journal of the Social Services Research Group – Vol 17 (2) 1999
Deprivation indexes: do they measure up?
David Cubey, Performance Improvement Manager, East Sussex Social Services Department
Of central concern to social services departments over the past 20 years has been how to allocate or target resources to those most in need of services. This has sometimes resulted in indexes of deprivation or indexes of health need, such as the Jarman index, being used for purposes for which they were not intended The focus of this paper is the identification of a valid and reliable model to allocate social services resources to the elderly population in East Sussex. The model operationalises the Breadline Britain index of deprivation (Gordon and Forest, 1995) at District level and combines the results with those obtained from operationalising a synthetic model of the levels of disability in the elderly, population (Opit, 1992a). It is argued that this model can be validly and reliably applied to other local authorities.
A central question faced both by central and local government is how most effectively to allocate resources to those most in need of health care or social services. A number of deprivation indexes, primarily based (but not exclusively) on census indicators such as the number of elderly people living alone and the number of people who are unemployed in an area, have been developed over the past fifteen years. Some of these have been used to target or to justify existing allocations of social services resources to geographical or operational areas within a larger local authority area.
The indexes that have been most commonly utilised in the past are the ‘Jartnan index’ also known as ‘Under Privileged Area Scores’ (Jarman, 1983) and the Department of the Environment (DoE) index of Local Conditions (1981 and 1991). More recently Gordon and Loughran (1997) have used a modified version of the Breadline Britain deprivation index developed from the Breadline Britain poverty survey (Gordon and Forest, 1995) to allocate resources for social services provision to children and families in Somerset.
Each index can be seen to reflect the different purpose for which it was intended, for example, the Jarman index (1983, 1984) was intended to measure the workloads of GPs in different areas so that resources could be allocated accordingly. Throughout:
A key concern is the targeting of assistance on the greatest concentrations of deprivation. (DoE, 1995:1)
The DoE index has been used to target resources as part of Urban Aid programmes. Townsend (1987) has utilised sociological methods and analysis both in offering a critique of deprivation indexes and in developing alternative models. The Townsend deprivation index has been used to demonstrate health inequalities between the south and north of England and Scotdep (Carstairs and Morris, 1991) has been used to demonstrate health inequalities in Scotland.
The Department of the Environment, Transport and the Regions (formerly the DoE) draws up Standard Spending Assessment formulae to determine the amount of money to be allocated to individual Local Authorities each year. However, there does not appear to be a clear conceptual basis to these formulae, rather they appear to be statistical regression models to arrive at a given financial allocation.
Opit (1992) has developed an alternative model for the allocation of health and social services resources. This relies not on levels of deprivation but on levels of disability in the population.
This paper identifies a model to allocate the East Sussex County Council social services budget for the purchase of services for elderly people from the independent sector across the five District Authority areas of Lewes; Wealden; Eastbourne; Hastings and Rother which make up the County Council area. The services purchased by the social services department (under the NHS and Community Care Act 1990) for older people over the age of 65 consists of domiciliary personal care to enable people to remain living in the community rather than enter residential care. Where this is not possible because the person is so frail and physically dependent that they need full time care then residential care is purchased. The greater part of the budget (65%) is spent on the purchase of residential care.
Because the Districts that make up East Sussex County Council are organised into operational areas, each with its own separate budget allocation for the purchase of services, it is seen as very important that the budget allocation reflects the relative levels of need for services in each operational area. The level of need for social services by older people in each area is related to the level of disability in the elderly population and the level of deprivation in those areas. The level of disability in the older population is significant because that affects the need for services or social care provision. The level of deprivation affects both the need for public services (as opposed to people purchasing their own services) but it also has an affect on health (Townsend, 1987) and hence disability.
A rational model for social services resource allocation allows for decisions for increases, or decreases, in service levels within a local authority to be informed by the levels of deprivation and disability in specific localities. This is an important counter to local and political lobbying where it is based on insufficient objective evidence.
Table 1 identifies the indicators at Ward level in the more commonly utilised indexes, so that direct comparisons can easily be made between them.
The question as to which is the most reliable and valid index was one posed by Lee et al. (1995) in their comprehensive review of deprivation indexes. The answer to this question is very dependent on whether the index is based on a sound conceptual framework. It also depends, not only on the variables included in an index and how the data are weighted, transformed, standardised and validated, but crucially on the intended purpose of operationalising the index.
The Breadline Britain index was deemed to be in a group of the most valid and reliable indexes in a review undertaken by Lee et al. (1995). A review of indexes by the University of York also concluded that the Breadline Britain index was the best available (Burrows and Rhodes, 1998). Both the Jarman and Breadline indexes attach weights to the data obtained on each variable. In the case of Jarman these weights are based on the perceptions of GPs as to what variables impact on their workloads. In the case of Breadline the weights were determined by the perceptions of a representative sample of the general population as to what constitutes a minimum standard of living.
The Jarman index was not operationalised for the purpose of this research since its stated purpose is to measure demand for primary health care. In private correspondence with Lee, Jarman (1994) acknowledges that the index was not intended to be a deprivation index as such (Lee et al., 1995). Ironically the Jarman index was found to have the lowest correlation out of ten indexes with standardised illness ratios and it was the least reliable out of all the indexes reviewed (ibid.).
|Children in flats||*|
|Long term illness||*|
|Low social class||*||*|
|Elderly living alone||*|
The DoE 91 index was deemed by Lee et al. to be amongst the least reliable and valid indexes. It appears, out of the range of deprivation indexes available, to be the most technically sophisticated. However, as the authors of the DoE 91 index acknowledge:
the index has been designed to identify relatively deprived areas. It is not well suited to saying which of two non-deprived areas is the most or least deprived. (my emphasis). This is because there are a great number of LA areas with little deprivation and therefore with similar scores. Hence small changes to their deprivation score could alter their rank position considerably (1995:5).
This would represent a severe limitation on using the DoE 1991 index in East Sussex as a way of allocating resources between Districts. Negative scores on the index indicate relatively low levels of deprivation. Four out of five of the Districts that make up the County Council have negative scores, with only Hastings having a positive score of ten.
Townsend (1987) criticised the previous DoE 1981 index on the basis that inner London local authorities tended to be ranked as highly deprived on a range of the variables used. These included overcrowding, households lacking standard amenities and to a lesser extent unemployment. The latter three variables are included in the current 1991 DoE index. Yet these London authorities were substantially lower down on a ‘league table’ of mortality levels compared with some authorities in the North of England.
Gordon and Forrest (1995) claim that the Townsend index is the most widely used index for mapping deprivation. However, it was in the mid range group of reliability and validity according to the Lee et al. (1995) analysis. Although it shares three variables with the Breadline index (no car; rented and unemployment) the latter index has additional variables (long term illness and low social class) which are considered to be specifically relevant to the allocation of social services resources to the elderly. The Breadline index is considered below.
The Breadline Britain index
The Breadline Britain index endeavours to estimate the percentage of households that are poor at Ward and District level. As Gordon and Forrest (1995) point out there were no questions on income in the 1991 census and this can be seen to be a major omission from the census. Gordon and Forrest:
estimated the percentage of poor households in each English district using a weighted index derived from the Breadline Britain in the 1990’s survey designed to measure the extent and nature of poverty- in Britain (Gordon and Forrest, 1995:6)
The Breadline Britain survey was conducted by Mori and defined poverty in terms of the public’s perception of the minimum resources that are required to participate in ordinary activities in society. People with fewer resources are more likely to experience deprivation. At a certain point below a certain threshold of resources, deprivation is likely to grow disproportionately to a further loss of resources. This threshold once crossed is seen as marking a state of poverty (Townsend, 1987,1993). The data obtained from the Breadline Britain survey were used to obtain weightings by logistic regression for the census variables that make up the Breadline index. The indicators used in Breadline are shown in Table 2 in the format used by Lee et al. (1995).
Table 2: Census variables for Breadline deprivation index
|No car||Number of households with no access to a car|
|Rented||Number of households not in owner occupation|
|Lone Parents||Number of lone parent households|
|Low social class||Number of workers in social class IV or V|
|Long term illness||Number of households containing a person with a long-term limiting illness|
|Unemployment||Number of people who are unemployed in the economically active population|
|Weighting||Estimated % of poor households (No car * 0.2174) + (Not owned 0.2025) + (Lone P * 0. 1597) + (Low class* 0. 1585) +(Long ill * 0. 1079) + (Unemp * 0.0943)|
|Validation||Breadline Britain poverty survey.|
|Standardisation||Standardised by estimated % of poor in each Ward.|
There is general agreement that unemployment is a fundamental measure of deprivation (DoE, 95; Carstairs and Morris, 1991). Although unemployment may be questioned as being relevant to the non-working elderly population, if an area has suffered unemployment for a considerable period of time then it will impact on pensions and assets generally.
The lack of a car is a proxy for low income and indicates a household’s accessibility to goods, services and jobs. It could be seen to be a more critical indication of deprivation in a rural area. East Sussex has a large proportion of the population living in rural areas. Non home ownership (rented) is another proxy for income and also can be taken as a reflection of wealth. Townsend et al. (1988) argue that, taken together with no car, it offers a good reflection of income levels. Long term limiting illness is a very relevant variable in relation to social services provision to older people. As Dale and Marsh point out:
the instructions on the (census) form specifically requested the inclusion of problems due to old age. (1993:35)
There is legitimate concern that there has been under reporting of long-term limiting illness by older people. However, Benzeval and Judge (1993) found that the census instructions appeared to encourage the reporting of a long-term limiting illness by those over 75 years of age (Dale and Marsh, 1993). Of more concern is the issue of coverage, with females over 75 and the widowed elderly being undercounted by government surveys (ibid.).
The long-term limiting illness variable will also be identifying illness in the non-elderly population. Nevertheless, as with being a lone parent, illness in the non-elderly population may significantly affect the support available to elderly parents from relatives or indeed even neighbours.
Payne et al (1996) found when undertaking a factor analysis on deprivation indicators, that social class (and unemployment) is an identifiable deprivation factor in urban, small town and rural areas. Lee et al. state that:
Breadline showed no evidence of double counting. (1995:54)
The Measurement of disability to allocate resources: ‘the Opit model’
An alternative to looking at the distribution of deprivation in the population as a way of allocating and targeting social services or health resources is to look at the levels of disability or illness in the population. One argument for doing this is that these are potentially direct measures against which to target resources, whereas deprivation could be seen as a proxy for illness and disability.
Opit undertook a statistical analysis of the OPCS survey of disability of the elderly in private residences as a basis for a synthetic model which it is claimed can:
predict the prevalence and severity of disability in a particular sub-population (1992:1)
A synthetic model takes certain characteristics of a sample of the population and then assumes that they apply to other samples of the population or to the whole population. Thus the 1985 OPCS disability survey measured a range of physical disability including: mobility; visual and hearing impairment; capacity for self-care and incontinence. An overall severity score was calculated for each person identified as disabled. The mean disability score of all people identified as disabled was then calculated for each of the 5 census age subgroups in the elderly population. The OPCS survey also estimated the prevalence of disability in each of the 5 age subgroups. The mean disability scores for each age subgroup are set out in Table 3, along with the percentage of people in each age subgroup who it is estimated will have some disability.
Table 3: Overall disability level and distribution by age groups amongst those with some disability (OPCS, 1985)
|Age groups||Mean score||% Some disability|
Opit (1992) concludes that age alone (if over 65 years of age and in a private residence) correlates with disability more than any other variable used in the Office for Population and Census (OPCS) ten year census. He advocates therefore multiplying the number of people in each age subgroup in any local authority area by the OPCS disability mean score for that age subgroup and by the disability prevalence rate for that age subgroup. If this is undertaken for each age subgroup giving a total weighted figure for a District this can be compared with a weighted figure for another District. Opit claims this is the most effective way of allocating resources to services for the elderly population between Districts and asserts that:
If we multiply the severity estimates (the disability mean scores) in each age group by the overall prevalence, we can then estimate the population rates… (1992:12).
Thus, Opit argues that a model for resource allocation can be constructed solely on the basis of 5 age subgroups of the elderly population as measured by the 1991 census.
What the Opit model does not take into account is the resources available to those with disabilities, even if in large samples the distribution of disability is constant. However, this latter point is questionable as Townsend et al (1988) demonstrated by showing the inequalities in health between the south and the north of England. Also, older people with high incomes may well purchase either health care or social care provision. For example, social services departments (SSDs) are required under Section 2 of the Chronically Sick and Disabled Persons Act 1970 to provide daily living equipment and adaptations to property where people are permanently and substantially disabled as defined by the National Assistance Act 1948. However, the provision of adaptations to property (e.g. the adding of an extension to provide a ground level shower and toilet where a person can no longer use stairs) is means tested in East Sussex and will not be provided when a person’s income exceeds a certain level. Differences in levels of income and assets between towns in a local authority (L.A.) area have implications for how resources are distributed or targeted within a L.A. Additionally Gordon and Pantazis point out that:
After allowing for age, deprivation and other poverty related factors …. have the greatest impact on disability (1997:50)
An analysis of the 1985 OPCS disability surveys Berthoud et al. (1993) indicated that 45% of all disabled adults were living in poverty.
On the other hand the provision of daily living equipment such as hoists and mobility equipment by East Sussex Social Services Department is not currently means tested. This may mean that income levels, or alternatively deprivation levels, will be less relevant than the levels of disability in the population as measured using Opit’s model. Nevertheless in areas where incomes (e.g. pensions) are higher, people may choose to buy their own equipment rather than have to wait for an Occupational Therapy (OT) assessment and then wait further for the supply of equipment. As in the NHS, social services departments have in the past had to operate waiting lists which has meant some people waiting years before a service has been provided. Thus there is a strong argument for allocating resources not only on the basis of levels of disability in the population but also in relation to levels of deprivation at an area level.
The final resource allocation model
In light of the above literature review a model was constructed that combines the Opit Model with the Breadline index. The levels of disability in the population correlate with age more than any other variable (Opit, 1992; Gordon and Pantazis, 1997).Levels of disability are one of the primary determinants of service provision by the SSD and hence the use of the Opit Model. Deprivation levels, as measured by the number of poor households (Breadline index), are used in the combined model because it is hypothesised they also affect the demand and provision of social services. Thus the model reflects levels of deprivation but is not the sole determinant of the budget allocation.
The Opit model
As already described, the Opit model uses estimated prevalence rates of disability for the 5 census age subgroups in the elderly population and disability mean scores (obtained from the 1985 OPCS Disability Survey).
The mean scores for each age subgroup are shown in Table 3 along with the percentage of people in each age subgroup who will have some disability.
The disability mean scores were applied as a weighting factor to the number of people in each of the 5 age subgroups in each District in East Sussex who have been estimated to have some disability (prevalence). The numeric values for each age subgroup are added together to give the District total which can then be used to describe the level of ‘need’. The numeric values for each District can be calculated as a percentage of the total for East Sussex. Thus the amount of need in each District is divided by the total amount of need (the sum of the Districts) multiplied by a 100. The total budget for elderly services is then multiplied by this percentage figure for each District to give the District budget allocation.
The resource allocation model combines the results from the Opit model with the results from the Breadline Britain index, using District level data as the way of operationalising the Breadline index. Table 4 below sets out the percentage of households that are estimated to be poor using District level figures for the Breadline index (People and Places 2, Gordon and Forrest, 1995). The number of household figures are 1996 estimates (East Sussex County Planning Dept). This allows the percentage of households estimated to be poor in each District to be recalculated as a percentage of the total of all households estimated to be poor (in the 5th column) in East Sussex. These percentage figures for each District are then used to allocate the budget. Thus Eastbourne with 23% of the total of households estimated to be poor in East Sussex would receive 23% of the total budget.
The OpitIBreadline Britain combined model
The results-of operational ising the Opit model are added to the results from the Breadline index using District level data (as in Table 4) and then divided by two to produce a final allocation for each District.
Table 5 shows the results of operationalising the ‘Opit model’ to allocate the social services department’s ‘spot purchasing’ budget for the elderly in East Sussex in the financial year 1996-97. Table 6 indicates the budget allocation using the Breadline index at District level. Table 7 summarises the results from the combined Breadline and Opit model. Figures 1 to 3 compare the budget allocations using the Opit model, the Breadline index at District level and the two combined, with the actual budget allocations in 1996-97.
Table 4: Percentage of poor households in East Sussex by district
|District||N of households||% of households poor||N of households poor||% of total|
Table 5: Allocation of spot purchasing budget for elderly people using the Opit model
|Area||1996-7 budget||% of budget 1996-7||Opit %||Opit budget||Gain/loss|
The allocation of the budget using the Opit model would have resulted in Eastbourne losing nearly £1m with Lewes and Wealden gaining two thirds (£666k) of the £1m and Hastings and Rother gaining approximately a third (£296k).
Figure 1: Allocation of spot purchasing budget for elderly people using the Opit model
[image not available]
Table 6: Allocation of spot purchasing budget for elderly people using the Breadline deprivation index at district level
|Area||1996-7 budget||% of budget||Breadline %||Breadline budget||Gain/loss|
It will be seen that Eastbourne would have lost £888k using the Breadline index at District level. The redistribution from Eastbourne using the Breadline index would have gone primarily to Hastings\Rother (£755k) and a small redistribution (from Eastbourne) to Lewes\Wealden (£133k).
Figure 2: Allocation of spot purchasing budget for elderly people using the Breadline deprivation index at district level
[image not available]
Table 7: Spot purchasing budget for elderly people – Opit model and Breadline deprivation index at district level
|Area||1996-7 Budget||% of budget||Opit/Bread %||Opit/Bread budget||Gain/loss|
As can he seen from Table 7 above, the final model combining the Breadline index (operationalised at District level) and the Opit model would have resulted in a nearly equal split in a redistribution from Eastbourne (£925k) to Lewes\Wealden (£400k) and to Hastings\Rother (£525k). Chart 3 below highlights the budget allocation resulting from this final model.
Figure 3: Spot purchasing budget for elderly people – Opit model and Breadline deprivation at district level
[image not available]
Analysis and discussion of results
The question arises as to why Eastbourne appears to have had so much over provision relative to other districts. As Table 4 demonstrates, Eastbourne is the second most deprived district with 20.2% households poor, only exceeded by Hastings with 21.5%. In absolute terms Eastbourne has the highest number of households deemed to be poor (having a higher number of households than Hastings). Nevertheless, it would appear that SSD managers in Eastbourne were able to secure a greater share of the NHS and CC Act budget in 1996/97 than justified by the Breadline/Opit allocation model. However, what this analysis does not show is the resource allocation to the SSDs direct ‘in-house’ provision of services to older people and logically this needs to be the next stage-of analysis.
The challenge for social service departments will be undertaking any realignment in resource allocation as a result of applying the rational resource allocation model. If resource allocation becomes more ‘scientific’ in the future this raises the question of the role of local authority councillors. However, the model does not deal with the question of what level of resources should be allocated to different categories of people receiving services within a district (e.g. people with learning disabilities versus older people). This is a clearly a political decision that will be influenced by competing stakeholders.
As can be seen from Figure 4, the final model combining the Breadline index and the Opit model sees nearly an equal split of the redistribution from Eastbourne (£925k) to Lewes\Wealden (£400k) and to Hastings\Rother (£525k). Thus the higher levels of deprivation in Hastings are balanced out by the higher population of older people in Lewes\Wealden (42 %) as opposed to Hastings\Rother (36 %). If the Opit model alone is applied it would result in two thirds of the redistribution going to Lewes/Wealden and only a third (£300k) gong to Hastings\Rother because of the higher population of older people in Lewes\Wealden.
Figure 4: Gains and losses by district using Breadline and Opit model
[image not available]
Finally it will be noted that there is little difference in the total percentage allocations to Districts using the Breadline index or the combined Breadline/Opit model (Tables 6 and 7). Thus Hastings/Rother would receive 39% of the budget using Breadline alone, as opposed to 38% using the combined Opit/Breadline model. Lewes/ Wealden on the other hand would receive 38% using Breadline but 39% using the combined model. However, this small differential in percentage allocations does significantly affect the pattern of redistribution of resources from Eastbourne to Hastings/Rother and to Lewes/Wealden. Applying the Breadline index alone would result in most of the redistribution from Eastbourne going to Hastings/Rother whereas applying the combined model results in a relatively even split in the redistribution away from Eastbourne.
To summarise, what does vary significantly in applying these two models is the pattern of redistribution of resources from Eastbourne to the other areas.
Validation against social services department data
This research was based on the financial year 1996-97. Tight controls were applied by East Sussex social services department managers to the expenditure of the NHS and Community Care Act (1990) budget in the following financial year (1997/98). Generally, only those cases in the highest eligibility banding for services were funded due to a projected budget shortfall. The result of applying these controls in effect centralised the budget instead of allocating it on a geographical basis according to the allocation in the previous financial year. The resulting percentage allocations for services to the elderly for each District matched exactly the percentage allocation indicated by the composite (Breadline/Opit) model. These data thus provide very strong independent validation of the resource allocation model advocated as a result-of this research.
Implications for resource allocation policies for SSDs
One of the key issues that faces the social services department in East Sussex and in other local authorities is the allocation of budgets to service provision that ensures equity to residents living anywhere in the local authority area. Thus, in East Sussex, somebody living in Hastings should have equal access to a service as a person in Eastbourne where their level of need for the service is the same.
East Sussex, along with other social services departments throughout the country, have generally allocated budgets according to the allocation in previous years, which could be seen as reflecting demand for services. However, demand in terms of usage of services may be as much a reflection of knowledge, accessibility and quality of the service as the level of need for the service. The need for a service provided by the social services department to older people on the other hand is likely to be a reflection of levels of disability in the population and the support available from all sources, including a person’s access to alternative support or services. Areas which have high levels of relative deprivation will generally have less people with alternative support or ability to purchase services such as domestic or personal help. There is also a correlation between deprivation and disability which is likely to be a two-way interaction.
This research indicates that the most reliable and valid model to use to allocate resources across Districts in East Sussex, for social services provision to older people, is a combined model of the Breadline deprivation index and Opit’s synthetic model of the levels of disability in the elderly population. It has only examined the position with respect to spot purchasing monies, provided under the NHS and Community Care Act (1990). Logically the model should also be applied to those services directly provided by the social services department,
This further analysis might show that an imbalance in the spot purchasing budget is offset by imbalances in directly provided services. Thus overall there may have been a more equitable allocation of resources in 1996/97 than indicated in this paper.
The Breadline/Opit model may also be the most appropriate model to allocate the budget between Districts for services for people with physical disabilities (which will include a large proportion of older people). For people with mental health problems and people with learning disabilities, the Breadline Deprivation index alone may be the most relevant model for budget allocation purposes and there needs to be further exploration of these propositions. In the case of learning disability such allocations may be complicated by the need to purchase very expensive residential care provision for individuals (e.g. at £50k per year) which could skew funding allocations if one area were funding more of such placements than another area.
What this research does not address is how resources should be allocated across categories of people needing services within a District. For example, what should be the proportion of the budget spent on older people as opposed to people with learning disabilities. This is largely a political question but research, such as reported in this paper, should be able to provide data as well as a conceptual framework to inform such decisions.
The other issue that has not been addressed is that highlighted in the social care White Paper: Modernising Social Services (DoH, 1998) as to the balance of expenditure on residential care for older people (approximately 65% in East Sussex) to home care. Central Government expects a shift from bed based care to home care. The rational resource allocation model proposed in this paper can provide a basis for any future reconfiguration of service provision.
The author wishes to thank Mr Chris Smaje, University of Surrey for his advice and comments during the course of the research; Dr Dave Gordon, University of Bristol for the provision of pre-publication material; and East Sussex County Council, particularly the valuable assistance of Sarah Boughton.
Arber, S. and Ginn, J. (1991) Gender and Later Life. London: Sage.
Berthoud, R., Lakey, J. and McKay, S. (1993) The Economic Problems of Disabled People. London: Policy Studies Institute.
Benzeval, M. and Judge, K. (1993) ‘The 1991 Census Health Question’, Discussion Paper, London: The King’s Fund Institute.
Bradford MBC (1993) Areas of Stress within Bradford District. Bradford: MBC, Chief Executives Department.
Burrows, R. and Rhodes, D. Unpopular Places? Area Disadvantage and the Geography of Misery in England. Bristol: Policy Press.
Carstairs, V. and Morris, R. (1991) Deprivation and Health in Scotland. Aberdeen: Aberdeen University Press.
Dale, A. and Marsh, C. (Eds.) (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.
DoH (1998) Modernising Social Services: White Paper. London: HMSO.
Forrest, R. and Gordon, D. (1993) People and Places. Bristol: University of Bristol, School for Advanced Urban Studies.
Frayman, H., Mack, L, Lansley, S., Gordon, D. and Hills, J. (1991) Breadline Britain In The 1990s: The Findings of the Television Series. London: Domino Films and LWT.
Gordon, D. and Forrest, R. (1995) People and Places 2. Bristol: University of Bristol, School for Advanced Urban Studies.
Gordon, D. and Pantazis, C. (Eds.) (1997) Breadline Britain in the 1990’s. Aldershot: Ashgate.
Gordon, D. and Loughran, F. (1997) Child Poverty and Needs Based Allocation, Research Policy and Planning, 15 (3): 28-38.
Jarman, B. (1983) Identification Of Underprivileged Areas, British Medical Journal, Vol. 286: 1705-1709.
Jarman, B. (1984) Underprivileged areas: validation and distribution of scores, British Medical Journal, Vol. 289: 1587-1592.
Jarman, B. (1994) Letter to Peter Lee outlining method of Underprivileged Score using 1991 census data.
Lee, P., Murie, A. and Gordon, D. (1995) Area Measures of Deprivation: A study of current methods and best practices in the identification of poor areas in Great Britain. Birmingham: Centre for Urban and Regional Studies, University of Birmingham.
Noble, M., Smith, G., Avenell, D., Smith, T. and Sharland, E. (1994) Changing Patterns of Income and Wealth in Oxford and 01dham. Oxford: University of Oxford.
Opit, L.J. (1992) Disability, Age and Social Context: Technical Analysis of OPCS Survey of Disability of the Elderly in Private Residences as a Basis for Synthetic Modelling. Canterbury: Centre for Health Services Studies, University of Kent.
Payne, G., Payne, J. and Hyde, M. (1996) “Refuse of All Classes”? Social Indicators and Social Deprivation, Sociological Research Online, Vol. 1 (1). <http://www.socresonline.org.uk>
Townsend, P. (1987) Life and Labour in London. London: CPAG.
Townsend, P. (1987) Deprivation, Journal ofSocial Policy, Vol. 16: 125-46.
Townsend, P. (1993) The International Analysis ofPoverty. London: Harvester Wheatsheaf.
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.