The Scottish Government

« Previous | Contents | Next »

Listen

Measuring Deprivation in Scotland: Developing a Long-Term Strategy: Final Report

5. Developing area-based measures

In recent years, there have been significant improvements in the availability of data for small areas and in the methodologies used to calculate area deprivation indices. The recently-published SIMD2003 takes advantage of both and represents a real advance. This section sets out a strategy for developing this index in the longer term. The Interim Report discussed at some length the set of indicators currently included in the SIMD2003 as well as possible additions, amendments and deletions. Here, we focus on presenting the conclusions we have come to. The justification for including each of the domains in the area index was discussed in Section 3.3 above. This section sets out in more detail the indicators on which we would base each of the domains. We distinguish between areas which can be implemented immediately or in the short term, and those where further development is required. We also set out our recommendations on a number of more specific issues.

5.1 Domains and indicators

5.1.1 Current income

There was some discussion during the consultation phase over whether there should be a single financial resources domain or whether current income should be kept separate. On balance, we favour keeping it separate as it is seen as such a key element in the overall index and it is based on comparatively robust data. This domain is concerned with measuring the proportion of people on low incomes as low income is the key cause of deprivation.

Benefits take-up

In the absence of reliable figures for local income distributions, the SIMD2003 used indicators based on the proportion of adults and children receiving "low income" benefits. We recommend retaining this approach. Some updating of the indicators will be necessary to reflect changes in the availability of benefits and tax credits but this will improve population coverage significantly. Specifically, Working Tax Credit (from April 2003) extends coverage of working age people without children who are in work but on low incomes, while the Pension Credit (from October 2003) will extend coverage of older people on low incomes but not in receipt of Income Support. In both cases, there will need to be an income cut-off to retain the focus on those on low incomes as is the case at present. Data would be obtained from DWP.

Benefits eligibility

There are concerns that take-up rates may vary across the country and that, as a result, they may provide a slightly biased picture of the distribution of low income households. We therefore recommend that further research be commissioned to examine spatial variations in take-up rates, building on earlier work by Bramley et al (2000). This work would be designed to produce correction factors to be applied to take-up rates so that the benefits data provides a more accurate picture of levels of eligibility and hence low income.

Local income distributions

In the longer term, there remains a concern that some low income groups may be missed by benefits data (many students, for example) and that this indicator is susceptible to changes in the benefits regime. Work is being carried out by ONS to model income distributions for small areas. While this does not appear to be sufficiently robust at present, it may provide the basis for direct estimates of the numbers of people below certain income thresholds in future. As this would remove all problems of population coverage/eligibility and take-up arising from benefits data, it would be preferable. We recommend that the Executive monitor the development of this data source. We also recommend that the Executive adds its voice to calls for Inland Revenue to play a more active role in making data on local incomes more widely available.

Other

There were mixed responses to the question of using Housing or Council Tax Benefit take-up. On the former, there was concern that this was available only to renters. For the latter, there was felt to be considerable overlap with other benefits, particularly now that the tax credit regime has increased coverage. While cross-checking of results might eliminate duplicate records, the added-value of such an exercise appears relatively small. We do not recommend incorporating either.

5.1.2 Other financial resources

The aim of this domain is to pick up additional factors which determine the level of financial resources available to an individual, beyond current income. There are three elements to this domain: persistent low income; savings and debts; and cost of living adjustment.

Persistent low income

There was strong support from consultees for devising measures of persistent low income. The most direct approach would be to use the low income benefits data and to seek to identify individuals who had been on one or other of these benefits for, say, a period of one year. As with low income, this could be expressed as a simple proportion of the population. There may be some difficulties where individuals have moved between benefits. We recommend that the Executive discusses the feasibility of obtaining such data with the DWP.

Savings and debts

The suggestion that measures of financial resources should include measures of savings/debts had widespread support. One additional argument in favour of this approach is that this would provide a better picture of the relative financial position of households over the life-cycle. Older people may have lower incomes but tend to have more savings to draw on. Significant challenges remain in terms of defining savings/debts and in terms of measuring these. On debt, a working group on over-indebtedness led by the Department of Trade and Industry (DTI) has drawn a distinction between secured debt (principally mortgage debt on housing) and unsecured or consumer debt (credit cards, bank overdrafts, consumer credit agreements, etc.). The latter seems the most appropriate area to focus on as it applies to the whole population although, for homeowners able to borrow by re-mortgaging, the line between the two may be very unclear. On savings, a similar distinction might be drawn between relatively inaccessible savings tied up in housing equity or pension funds and those more readily accessible in savings accounts or short to medium terms investments.

For debts, one approach is to measure levels of debts directly. These should be related to income levels as it is not debt itself but the cost of repaying debt in relation to income which matters. There are no public sources of data at present. Some commercial data on credit-worthiness exists and this could provide indicators immediately and on an on-going basis. Questions about the validity of the data would need to be addressed. The alternative approach is to look at experiences of financial stress (problems paying bills, arrears of mortgages or rent, reduced consumption of necessities for financial reasons, etc.). Data on these issues could be gathered through household surveys and modelled to small area level. In the short term, we recommend that the Executive explores both the purchase of data from a private supplier and the possibility of modelling data.

In the longer term, alternative sources of administrative data may become available. There are moves to develop a common system for money advice or debt counselling to be used across public and voluntary sector agencies. While this would only provide data on those who approached such agencies for help, the information would be available on a consistent basis. We recommend that the Executive monitors these developments.

On savings, some very basic data is collected in the Scottish Household Survey which could be modelled to small area level although the accuracy of this would need to be checked. There may also be commercial data. Again both options could be explored further.

Cost of living adjustment

There was strong support, particularly from rural areas, for the inclusion of some form of measure of relative cost of living. We are aware that it might prove very expensive to collect the necessary data if a standalone exercise were required. Some data is collected from different locations for the existing Retail Prices Index. We recommend that the Executive examines whether this might be used to provide estimates of local cost of living for the whole country.

5.1.3 Housing

The argument for including a housing domain is that this is one area on which direct measures of living standards are available for small areas although, as noted above, the relationship between housing conditions and deprivation is less direct due to direct state provision (social housing) and additional financial support (Housing Benefits). We recommend that the domain should focus on direct measures of inadequacy of housing covering physical conditions, suitability and security of tenure.

In the short term, the Census can provide valuable if limited indicators on the physical conditions and suitability. We recommend the following indicators:

• the proportion of people living in housing lacking amenities and/or lacking central heating;
• the proportion of people overcrowded, defined as those living at levels above the occupancy norm.

This data cannot be updated directly but better measures of physical conditions and suitability should soon be available from the Scottish Neighbourhood Statistics work on housing. Assuming this links to the work to develop an updated minimum standard of housing (HITF, 2003), this would provide a much more satisfactory basis on which to measure poor physical conditions. We also recommend that the data on origins of homeless people seeking accommodation from local authorities be examined when it becomes available.

Some respondents to the consultation process favoured the development of measures such as housing costs or affordability (cost/income). Others argued that, if there are significant affordability problems, they would be expected to lead to more direct outcomes such as overcrowding or homelessness and that it should be these impacts which should be measured. We favour this approach.

5.1.4 Health

This domain is well developed in the current index and we would retain it with minor changes. We would, however, suggest that it is re-named "health" to reflect the focus on mortality and morbidity, rather than disability. In Section 3.3, we argue strongly that the justification for this domain is that it measures health inequalities associated with low income and deprivation. As such, the aim should be to identify areas with higher than expected levels of ill health or mortality given the age/sex profile of the population.

Standardising indicators

To this end, it is necessary to ensure that the indicators focus on this excess by controlling for age/sex differences as appropriate. Two indicators (Comparative Mortality Factor (CMF) and Comparative Illness Factor (CIF)) would be adopted unchanged as these have adjusted raw mortality or sickness rates to take account of population age/sex structures. Further work should be carried out on the other five indicators to explore whether population structure has a significant effect on these and to consider controlling for this if necessary.

Comparative Mortality Factor

We would extend the measurement of mortality, CMF, to those over 75 to provide better population coverage. The relationship between deprivation and health is not as strong with older age groups but it is still significant and this would provide better coverage of the whole population. Given comments below about disaggregating indicators to distinguish between different age groups, this might be done by constructing separate indicators for working age and older people in the first place.

Comparative Illness Factor

The Comparative Illness Factor (CIF) is currently based on data from DWP on uptake of certain benefits because data on self-assessed health from the 1991 Census was considered too out-of-date. The limitation of this data is that it provides a rather uneven measure of ill health. People of working age are recorded as "ill" if they require help with personal care or mobility (Disabled Living Allowance) or are unable to work due to ill health (Incapacity Benefit or Severe Disablement Allowance). People over working age are recorded as suffering ill health if they require help with personal care (Attendance Allowance). Benefits measures have the attraction that they provide comparability between areas but they are not a useful guide to the absolute scale of problems or to the relative differences between age groups.

With publication of the 2001 Census, data is available for small areas on the number of people who see themselves suffering from a limiting long term illnesses (LLTI) and the number who rate their general health as poor in the past year. Some different views were expressed about whether self-reported health is a good guide to actual morbidity but it is clearly preferable to benefits-based measures on the basis of its coverage. We recommend that the Executive adopts both Census measures in the short term in place of the current benefit measures. Given that age has a strong impact on self-reported health, it is important that the Census measures are standardised in the same way as at present.

In the longer term, the SHS collects some data on self-reported health problems and this should be used to update Census figures in a process similar to that already used to estimate the proportion of working age adults lacking qualifications.

Life expectancy and healthy life expectancy

Work is underway within the Information Services Directorate of NHS Scotland to develop small area estimates of life expectancy and healthy life expectancy. These might provide a useful alternative to mortality and morbidity indicators, not least because they could be expressed as a proportion of the population (e.g. proportion with below average life expectancy or healthy life expectancy). We recommend that the Executive monitors developments with this data.

Smoking rates

There was strong support for adding in measures of smoking rates, given that these are seen as the major preventable cause of premature death and that there is a strong social class gradient. Some work to model small area smoking rates has been commissioned for Scotland by NHS Health Scotland (formerly the Public Health Institute of Scotland). This is using data from the Scottish Health Survey combined with Census population figures. We recommend that the Executive await the results of that work and considers including the indicator developed.

5.1.5 Education, skills and training

As discussed in Section 3.3, we recommend that the focus of this domain should be on educational attainment as an indirect measure of deprivation, both as an outcome but also as a cause.

Children and young people

For the school exam performance data, we recommend one minor change: using an average figure for three consecutive years to reduce the impact of fluctuations from one year to the next. We recommend retaining the indicator on the proportion not staying on in full-time education (either at school and or in higher or further education). Following feedback from the consultation process, we recommend that the indicator on absence rates be retained. The construction of the indicator can be significantly improved as pupil-level data on absence rates becomes available from the development of educational statistics.

Working age lacking qualifications

We would retain the indicator on working-age adults with no qualifications although, in the short term, we would replace the current measure with one based directly on the 2001 Census question on this issue. In the longer term, we recommend that the Executive should re-examine the method for updating small area estimates using LFS data. The current method relies on a model derived from data on the 32 local authorities and appears rather crude. A better alternative would be to develop the model on the basis of individual data from the SHS. (A separate note is being submitted to the Executive on this point.)

Poor literacy or numeracy

To strengthen coverage of education for the adult population, we recommend that a further indicator be developed to measure problems of poor literacy or numeracy. In line with the aim of this domain, this is intended to act as an indicator of underlying problems of deprivation although we expect that there will also be interest in the data for their own sake. The indicator will need to be based on individual data and modelled to small area level. At present there is no suitable data collected on an on-going basis other than as part of the International Adult Literacy Survey and the definitions used in that survey have been questioned (Scottish Executive, 2001). There is a strategy to tackle problems of poor literacy and numeracy being developed within the Executive and it is recommended that development of suitable indicators be conducted in conjunction with that work.

5.1.6 Employment

In our view, the employment domain does not fit within the conceptualisation of deprivation we have outlined and should therefore be dropped from the index. We recognise, however, that there was strong demand from many people consulted to retain this domain as unemployment is seen as a key issue in its own right. We therefore provide a brief discussion of how the indicators in the current index might be updated.

One issue which needs to be resolved is whether the domain attempts to follow the current justification, measuring "enforced exclusion from the world of work" (Noble et al, 2003: p.14) or whether its function is to act as a proxy for low income and hence an indirect measure of deprivation.

The latter approach is consistent with the conceptual framework set out above and would be our preference. If this approach were adopted, the measure would simply need to capture all those of working age who were not in (full-time) employment as a proportion of all of working age. Reasons for not working (full-time) would be irrelevant. Data on employment status is available from the Census and the measure could be constructed as the opposite of the (full-time equivalent) employment rate. This could be updated using data from the Labour Force Survey and/or benefits data or (if available) estimates of the employment rate taken from SNS. There would be substantial overlap with those recorded as claiming out-of-work benefits in the current income domain but this measure would also include those not eligible or not claiming benefits.

If the existing approach is retained, it is necessary to measure who is not working at present but would be working if: there were more demand for labour in the area; or they did not have health problems or disabilities which prevented them working. The current measure includes those in receipt of Job Seekers Allowance or on compulsory New Deal programmes, and those in receipt of benefits related to incapacity to work on grounds of health or disability (Severe Disablement Allowance or Incapacity Benefit). This should be retained. One problem with this measure, however, is the undercounting of who is "unemployed", particularly women. With unemployment benefits, there is no requirement for (usually female) partners of JSA claimants to sign-on themselves. Some of these partners may make a positive choice not to seek work, perhaps because they take on the unpaid work of caring for children or others. For other partners, the choice is at least partly determined by judgements about conditions in the local labour market and the type of work they might hope to obtain. While this bias is unlikely to affect the relative position of different areas, it does lead to an undercount of the absolute scale of the problem and to a bias against women in the index. Another problem with the concentration on unemployment is that the measure does not pick up those working part-time who might wish to work longer hours. Women are far more likely to work part-time than men.

Whichever approach is adopted, we recommend that any measure based on benefits data use figures for four successive quarters in order to remove problems of seasonal fluctuations or bias.

5.1.7 Crime and social order

The basis for this domain is that it identifies a range of matters which are seen as important problems or area deprivations in their own right. This domain may also be seen by some as an indirect measure of individual deprivation as well although that is not its role here. The argument for this is that, at an area level, there is good evidence of an association between some types of crime such as burglary and other aspects deprivation (Johnson et al, 1997; Smith et al, 2001). It should be noted that this does not mean that the poorest necessarily suffer much higher victimisation rates. Pantazis and Gordon (1998) provide evidence that it is "better off" households in poorer areas who are most likely to be victims of crime but it is the poorest who have much greater fear of crime. The domain therefore includes crime incidence and fear of crime, as well as a wider range of low-level problems associated with social disorder or "incivilities" such as vandalism, graffiti and noise.

There was strong support for including this domain from the consultation process, particularly from those concerned with urban areas. Some respondents from rural locations felt that problems of crime were not strongly associated with problems of deprivation in their areas. If this is the case, this makes crime measures a biased indicator of levels of individual deprivation but does not mean that they should not be used as indicators of problems in their own right.

Crime incidence

Several problems exist with using police recorded crime statistics to measure crime incidence, including variable reporting rates by the public, variable recording practices by the police and their unwillingness to make data available for small areas. In the short term, we recommend that the Executive adopts the approach to estimating small area crime rates which has been proposed for inclusion in the English deprivation index by Robson et al (2002). This apportions police recorded crime figures to ward level using a combination of victimisation rates from crime survey data and Census data on ward characteristics. The method was tested in a number of case study areas in England where local surveys provided direct estimates of crime incidence and it appeared robust. Further testing might be carried out in Scotland.

For the longer term, we recommend that the Executive supports efforts to introduce more standardised recording practices and explores further with police forces the possibility of making existing data available for small areas directly. Correction factors would need to be developed to allow for differences in recording levels between areas. These could be derived from an analysis of crime survey data in a process similar to that used to adjust benefits take-up rates to estimate eligibility (see Section 5.1.1 above).

Fear of crime

Using data from the Breadline Britain survey, Pantazis and Gordon (1997) argued that fear of crime is more strongly related to poverty than incidence of crime itself at an individual level. One important factor explaining this link is that many people living in poverty cannot afford household insurance so the consequences of burglary for them may be particularly severe. Data on fear of crime is available from the Scottish Crime Survey and would need to be modelled to small area level.

Social disorder and "incivilities"

Data on perceptions of problems of social disorder would capture some of the low-level crimes which are not recorded in crime statistics such as vandalism or graffiti, as well as nuisances or incivilities such as noise from neighbours, disturbance from passers-by and so on. Data is available from the Scottish Crime Survey and the Scottish Household Survey and this would need to be modelled to small area level.

Fire service data

The SIMD2003 report suggested that fire service data on malicious fires or false call outs might provide one indicator for levels of social disorder and this had some support in consultations. It is not clear whether it would be subject to the same problems of differential recording practices that apply to police recorded crime data. In the consultation process, it was suggested that this data would provide information on only a very small proportion of the types of crime impacting upon local areas. Nevertheless, we recommend that this data should be requested from Fire Services and examined to see how robust it is.

5.1.8 Physical environment

There was strong support for the inclusion of a domain capturing quality of the physical environment as this is seen as being important in itself. It may also be seen as important where the quality of the physical environment has an impact on health or well-being and hence, indirectly, on deprivation. In the Interim Report, we suggested that this domain should focus on negative features such as graffiti, litter or vandalism but we now include these within the crime and social order domain. We now argue that the focus should be on aspects such as air, water or noise quality, as well as proximity to positive or negative features (such as green space, mobile phone masts or noxious industries).

The Scottish Centre for Infection and Environmental Health (SCIEH) is carrying out work for the SNS project to develop a range of indicators of environmental quality. These cover: air and water quality; radiation levels; location of mobile phone masts, cooling towers and certain industrial processes; and complaints about noise. These should provide a useful source from which to develop a range of indicators and the Executive should look to build on this work when the indicators become available. The reliance on complaints about noise may be problematic. As different groups may have different propensities to complain, it may be a poor guide to levels of noise in an area. The relationship between this and other indicators of deprivation should be checked before it is included in the index.

There were a number of suggestions that this domain should capture proximity to open space or green space which could be used for recreational purposes. Much land around small settlements in rural areas may look attractive but could not be used for leisure or recreation. Different types of area would be needed for very young children, older children or adults. It would be relatively easy to collect data on the location of public parks or play areas for children and to calculate proximity to this using a simplified form of access measure (ideally based on walking distance). We recommend that the Executive explores the possibility of local authorities providing data on the location of such facilities. It is rather more difficult to define and measure access to other forms of open or green space which are much more variable in nature. We make no recommendations in that regard.

5.1.9 Geographic access and telecommunications

The justification for this domain is that it captures a set of problems which operate at area level and which are seen by many as important problems in their own right. These are the problems (financial cost, time and inconvenience) of having to travel a relatively long distance to access basic services. From this perspective, the fact that the domain has a negative correlation with indicators of individual deprivation in the current index is not relevant.

The domain is not intended as an indirect measure of poor access as a cause of deprivation due to its impacts on increased costs of travel and hence cost of living. First, higher costs in accessing services might be offset partially or wholly by lower housing costs. Second, if there is a more general cost of living problem, this should be captured directly in the cost of living measure proposed in Section 5.1.2.

Geographic access to services

The current domain focuses on geographic access to a number of key services. We recommend some changes to the list of services and the method for calculating distance, and the inclusion of access to employment.

We recommend that the focus should be on services which are visited regularly and are widely used, particularly by those on low incomes. We recommend retaining access to doctors, supermarkets, petrol stations and primary schools but dropping access to a bank/building society in favour of access to a post office. According to Bramley and Ford (2001), post offices are used by the vast majority of people and are seen as an essential local service. It is also likely to be a service of particular importance to many on low incomes. Data on location of post offices should be readily available. Access to a community internet facility should also be dropped as these are used by such a small proportion of the population and many alternative sources of access to the internet exist (home, work or commercial facilities). There were various suggestions of services which might be added, including some support for including access to secondary schools.

In terms of measuring distance, feedback from the consultation process was clearly of the view that drive or travel times rather than road distance would be preferable. This would also provide a better reflection of the difficulty of accessing services where ferry journeys were involved. From a policy perspective, the indicator would also show improvements both where new services were located closer to residents and where improvements in roads or transport services reduced travel times. This change should be made relatively quickly

Two further issues raised during consultation were the possibility of reflecting the availability of public transport services and the financial as well as time costs involved in making a journey. One means of tackling both issues would be to use "generalised travel costs". These reflect physical distance but also: type of transport available to an individual; time taken (including time waiting for connections); and convenience or quality of transport. Hence, two neighbours might have different generalised travel costs to reach the same destination if one has a car available to make the journey and the other has to travel by public transport. For the area as a whole, the cost of reaching a given destination (GP, shop, etc.) would be the average of the cost of getting there by car and the cost of using public transport, weighted by the proportion of the population relying on each mode.

Geographic access to work

There was wide support for including a measure of access to employment in the index. There was some debate about whether this was best included in the employment domain or the access domain. On balance, we feel that the issues involved (of transport, travel cost, location) mean that this is best included in the access domain.

In measuring access to employment, there is the additional complication that jobs are "rival goods" in the sense that only one person can occupy each post at any one time, as noted in the Interim Report. There is therefore a need to include distance of rivals (other potential job seekers) to the relevant jobs as well as distance of the individual to the workplace in question. The measure also needs to reflect the fact that access to a job depends on other factors such as level of skill.

A range of methods for calculating geographical access to employment have been developed, although the most sophisticated are those of Shen (1998). These allow for the incorporation of residential location, employment location, competing workforce, occupational differences and differences in mobility. They are quite demanding in terms of data requirements but could be operationalised at small area level using the Census. Some of the data could be updated between Censuses. Workplace employment data is available for small areas from the Annual Business Enquiry and local authority-level data from the Labour Force Survey could be combined with small-area population figures to provide some updating of workforce characteristics.

Telecommunications infrastructure

This had been proposed in the Interim Report as a possible additional domain but feedback suggested in might be included within the access domain as it captures a very similar set of problems. Rather than focus on individual ownership of particular items, it was felt that the indicators should measure the extent to which it was possible to gain connection to digital means of communication at reasonable cost. This could be achieved by measuring coverage of telecommunications infrastructure: mobile phone networks, broadband access, cable television or digital radio transmissions, for example. The indicators would be the proportion of the population able to gain access to each. Data would be obtained from service providers as there are no issues of confidentiality involved or from industry regulators.

Other issues in access to services

The Interim Report raised a number of additional issues which might have been considered in an expanded domain on access to services more generally. This would have covered issues such as:

• the quality and quantity of services being accessed (quality of school, waiting time to visit GP, for example); and
• the extent to which costs (charges as well as travel costs/time), quality or location of services acted as barriers to use.

There was some support for developing these areas but no further suggestions on how these might be measured. There are very significant challenges in measuring service quality or in assessing quantity of service provide in relation to levels of need. Data on barriers to service usage might be gathered through household surveys. The PSE1999 employed questions on the extent to which people were: using a service despite perceiving it as inadequate; not using a service because it was unavailable or inadequate; and not using a service because they could not afford to (Bramley and Ford, 2001). Modeling results to small area level might be regarded as relatively unsatisfactory in this case as a key determinant is expected to be variations in local provision. As a result, we recommend that the Executive keeps this area under review.

5.1.10 Social relations and social capital

The Interim Report considered how a domain might be constructed to capture social dimensions of individual deprivation on the basis that this would provide a direct measure of an aspect of deprivation not well developed in previous area indices. There was some discussion on which aspects of social relations should be captured and a contrast was drawn between social relationships as a source of personal support ("getting by") and as a source of contacts, connections or information ("getting on"). We also discussed the broader concepts of social capital, social equality or stigma and social order although the relationship between these and individual deprivation was less clear.

In this report, however, we have drawn a distinction between those domains which attempt to measure deprivation directly (e.g. income, housing, health and education) and those which are intended to capture area-wide problems of deprivation (e.g. crime and social order, physical environment or geographic access). The former are intended to provide coverage of all aspects of personal deprivation, both material and social. The latter are measures of area characteristics which are seen as related to or affecting problems of deprivation. It is difficult to justify the inclusion of a domain on social relationships as a direct measure of deprivation, not least it is difficult to envisage what sources could be developed to provide this information directly. This domain is therefore included on the basis that it picks up a number of the social dimensions of an area relevant to the understanding of deprivation. These are certainly seen as important issues in their own right, as current policy statements make clear (Scottish Executive, 2002a) and there was widespread support for including them from the consultation process.

The Office for National Statistics (ONS) has been leading on a project to develop small area data on social capital since 2001 and this is nearing completion. The SNS is already linked in to this work. According to the ONS website, social capital is being defined using an OECD definition as "networks together with shared norms, values and understandings that facilitate cooperation within or among groups" (Cote and Healy, 2001: p.41). There are seen as being five dimensions which the measure should capture:

• participation, social engagement, commitment;
• control, self-efficacy;
• perception of community level structures or characteristics;
• social interaction, social networks, social support; and
• trust, reciprocity, social cohesion (National Statistics, 2002).

We recommend that the Executive continues to monitor the development of this data and considers including the measures within the area index.

5.2 Units for analysis

The move from postcode sectors to wards in the SIMD2003 has been widely welcomed but there are problems with these units in the longer term. First, for some people, particularly but not only those in rural areas, wards are still too large and therefore miss important "pockets" of deprivation. Second, the possibility of proportional representation leading to major changes in wards, even an end to wards, needs to be taken into account. It was also pointed out that the two tier approach we had suggested in the Interim Report (a full set of indicators at ward level, a sub-set for the smaller data zones) would not be feasible as data zones do not nest within wards and problems of "differencing" would therefore arise.

We recommend that wards be retained for the immediate future but that the Executive looks at using the small data zones in the longer term. If there are problems with quality of data for the small data zones, then a two-tier approach might be considered where the upper tier is made up of aggregations of small data units.

5.3 Domains versus social groups

In the Interim Report, we set out a very different way of combining any given set of indicators to construct an overall index. Rather than grouping these into domains based on functional distinctions (income, employment, education and so on), it was proposed that indicators should be grouped according to age or "life-stage" (children, young people, working age and so on). The arguments for doing so were partly based on a criticism of the domain approach (notably that it was wrong to think of the education domain as highlighting educational problems, and so on). But it was partly based on a positive argument that the life-stages approach would provide information which might prompt more joined up thinking. A table was included in the Interim Report which showed how the current set of indicators could be re-cast in this way, with little or no change to the information collected.

There was significant support for both approaches to be used in future. There are only minimal implications for data collection, most notably the need to ensure that indicators can be disaggregated by age group. For the current set of indicators, for example, this would apply to data on low incomes derived from benefits and to data on adult health problems, derived primarily from NHS data sources. We recommend this approach be adopted.

In addition, there was also demand for the indicators to be disaggregated on the basis of other factors, notably gender. We recommend that the Executive seeks to provide gender breakdowns for all indicators.

5.4 Change over time

One specific issue on which we were asked to comment was the possibility of using area deprivation indices to measure absolute changes over time. As already noted, one of the limitations of the area based approach is that is a relative measure. The scores which each area achieves do not reflect the absolute number of people who are multiply deprived. An area may see a reduction in absolute levels of deprivation but no change in its relative ranking and vice versa. There was strong support from consultees for absolute measures of change.

In practice, the best that could be achieved would be to summarise change over time in terms of a "basket of indicators". In other words, absolute change on each underlying indicator could be published alongside the change in the relative ranking. For example, an improvement of ten places in overall ranking might be accompanied by a fall of 10 per cent in the number of low income households, a fall of 5 per cent in absence rates, a rise of 7 per cent in emergency admissions, and so on. In general, there was strong support from consultees for the Executive publishing more detail for individual indicators, rather than relying on the five domain scores as at present.

5.5 Updating the area-based measure

Our task is to set out a long-term strategy for developing measures of deprivation in Scotland. No matter how far-sighted this work tries to be, there will always be the need to review the measures. There may be opportunities to include new indicators, including those arising from the development of additional data sources recommended as part of this work. There may be methodological advances. There will be changes in standards of living and in social judgements as to what constitute "necessities of life". This section briefly considers some of the issues which arise for updating the area-based measures.

There is a trade-off between retaining a consistent measure to enable analyses to be made of change over time and updating a measure to take account of new data sources, methodological advances or changes in social perceptions or practices. Area-deprivation indices have rarely retained the same method for successive measures, being updated as new data sources and methods have become available. Up to now, however, these measures have only been produced on an occasional basis (following the decennial Census, until recently) so consistency seemed less important. With the move to producing deprivation indices on a more regular basis, possibly annually, a compromise position might be useful. This would entail retaining the same measure for a period of several years. Several revisions could then be made at one time rather than minor changes made each year. By publishing both old and new indices in the change years, it would be possible to distinguish changes in ranking due to change over time from those due to the change in indicators. A more continuous picture of the changing pattern of deprivation could then be developed.

One specific issue relates to the way in which individual indicators are combined to produce an overall domain score. For three of the current domains, the SIMD2003 uses factor analysis to derive weights. This has been criticised both for producing results which lean heavily on one or two indicators in some cases (such as the Education, Skills and Training domain) and, more importantly, for producing results which are unstable (Raab, 2003). This last issue is particularly problematic when it comes to updating the index. If factor analysis is repeated each year and this produces quite different weights, movements in the ranking of areas may owe more to changes in weights than to changes in local conditions. On the other hand, if weights from the first year are used regardless of subsequent changes in the relationships between variables, this might also be seen as unfair. We recommend using a consistent set of weights from year to year, revising these periodically at the same time as new data sources are included. We also recommend that the Executive reconsiders whether factor analysis is the most appropriate means of arriving at these weights.

5.6 Other issues

Our remit in conducting this review of the area-based measure specifically excluded consideration of technical issues. Several respondents argued that it was difficult to make a hard distinction between these and we have already commented on the use of factor analysis to weight indicators within each domain. One further issue relates to the impacts of shrinkage techniques on the picture of deprivation in rural areas where ward sizes were particularly small. There was concern that this added to the problems of identifying areas of deprivation in rural contexts, as potentially deprived areas had their scores "shrunk" towards the average for their authority. These problems would presumably become even greater if there were a move to use smaller data units in future. We have not examined the scale of these problems in practice but would note that it ought to be possible to construct more sophisticated shrinkage techniques. At present, scores for small areas on one variable are shrunk towards the local authority average for that variable. A more sophisticated approach would take account of the scores achieved by other small areas which were similar on the basis of other characteristics.

« Previous | Contents | Next »