Scottish Government

Scottish Indices of Deprivation 2003

Chapter 2: Domains and Indicators

Following on from the conceptualisation of multiple deprivation outlined in Chapter 1 the new Scottish Indices of Deprivation comprise indicators which are combined to form domains of deprivation. This process produced a score for each of the domains - a Domain Index - which has been ranked across Scotland to give a relative picture of each dimension of deprivation. The Domain Indices were then combined into an overall Scottish Index of Multiple Deprivation (SIMD).

An introduction to Domains and Indicators

Domains

The domains in the Scottish Index of Multiple Deprivation are Income Deprivation, Employment Deprivation, Health Deprivation and Disability, Education, Skills and Training Deprivation, and Geographical Access to Services.

Each Domain is presented as a separate Domain Index. Each domain reflects a particular aspect of deprivation. Thus the Employment Domain captures exclusion from the world of work and conditions of work - not the low income that may flow from it. The Income Domain can be used separately from the SIMD to examine low income alone. The Education Domain represents educational disadvantage and does not include markers of income deprivation such as 'children in receipt of free school meals', as children living in low income families are measured within the Income Domain. This approach avoids the need to make any judgments about the complex links between different types of deprivation (for example the links between poor health and unemployment), and enables clear decisions to be made about the contribution that each domain should make to the overall SIMD.

While the domains represent distinct dimensions of deprivation, it is perfectly possible, indeed likely, that the same person could be captured in more than one domain. So, for example, if someone was claiming Income based Job Seekers Allowance and had no qualifications, they would be captured in both the Income and Education Domains. This is entirely appropriate because one individual can experience more than one type of deprivation at any given time.

Indicators

Each Domain Index contains a number of indicators. The criteria for these indicators were that they should be:-

• 'domain specific' and appropriate for the purpose (as direct as possible measures for that form of deprivation)
• measuring major features of that deprivation (not conditions just experienced by a very small number of people or areas)
• up-to-date
• capable of being updated on a regular basis
• statistically robust
• available for the whole of Scotland at a small area level in a consistent form

The intention was to include a parsimonious collection of indicators that comprehensively captured the deprivation for each domain, within the constraints of data availability.

The indicators that were included in the Scottish Indices of Deprivation have been constructed using a range of techniques. Some of the data were obtained at individual level (with due regard to issues of confidentiality) and aggregated to ward level; some were obtained at other levels (e.g. Labour Force Survey Local Authority data) and then 'modelled down' to ward level. Postcoded data were assigned to 1999 wards using a postcode lookup table supplied by GROS. The assumption had to be made that postcodes supplied were correct and accurate and they were therefore used as given. As far as possible, all the data included in the Indices relate to April 2001.

The small numbers problem and the shrinkage technique

One problem which had to be addressed at the outset of the construction of the SIMD was the question of how the indicators should be scored or scaled (if at all) to allow fair comparisons between areas and appropriate combination with other indicators. The data were not all in the same units of measurement and if the raw data had been used the results would have largely been driven by the size of the population. For these reasons it was not possible to count the numbers of people experiencing each deprivation and add them together. Instead where possible, rates, or some other standard form of measurement were used which allow areas of different sizes to be compared.

In some areas of Scotland, particularly where populations at risk are small, data can be unreliable with particular wards getting unrepresentatively low or high scores on variables in certain domains. The extent of a score's 'unreliability' can be measured by calculating its standard error.

This problem emerged in the construction of other Indices of Deprivation in the past and this has prompted the use of the signed chi squared statistic (see DETR, 1998; NISRA, 1994). However, this technique has been much criticised for its use in this context because it conflates population size with levels of deprivation (Connolly and Chisholm, 1999). Given the problems with the signed chi squared approach, another technique - 'shrinkage estimation' - has been used subsequently to deal with the problem (see Noble, Smith, Wright et al, 2000; Noble, Smith, Penhale et al, 2000; Noble, Smith, Wright et al, 2001).

Shrinkage involves moving 'unreliable' ward scores (i.e. those with a high standard error) towards another more robust and appropriate figure e.g. the mean score of the local authority within which the ward is located. This may be towards more deprivation or less deprivation.

The actual mechanism of the procedure is to estimate deprivation in a particular ward using a weighted combination of (a) data from that ward and (b) data from another more robust source (for example the local authority mean). Using this method the estimate for any ward would then, for example, move towards the local authority mean by taking a weighted average of the ward and local authority values, thus reducing any ward-level 'noise' caused by small numbers. By this device the unreliability of the ward-level indicator is reduced by 'borrowing strength' from a more reliable source thus minimising the effect of random fluctuations and other sources of error. This methodology has a sound statistical basis and avoids the problem of indicator values being linked to the size of the area (scale dependency).

Although all scores move a fraction, only 'unreliable' scores, that is those with a large standard error, move significantly. The amount of movement depends on both the size of the standard error and the amount of heterogeneity amongst the wards in a local authority. The shrinkage procedure and formulae are presented in more detail in Appendix 1.

Combining the indicators into Domain Indices

For each domain of deprivation (Income, Employment, etc.) the aim is to obtain a single summary measure whose interpretation is straightforward in that it is, if possible, expressed in meaningful units (e.g. proportions of people or of households experiencing that form of deprivation). In some domains (i.e. the Income and Employment Domains) where the underlying metric is the same and where the indicators are non overlapping the indicators can be simply summed. Where there are several indicators within a single domain that have different underlying metrics and cannot therefore be straightforwardly combined (i.e. the Health and Education Domains), a statistical procedure, factor analysis, can be used to identify weights for each indicator. Factor analysis was also applied to the Access Domain. The domain score is then a combination of the component indicators weighted according to the factor analysis results. For further details on factor analysis see Appendix 2.

Income Deprivation

Income Deprivation: Indicators

• Adults in Income Support households (DWP, April 2001)
• Children in Income Support households (DWP, April 2001)
• Adults in Income Based Job Seekers Allowance households (DWP, April 2001)
• Children in Income Based Job Seekers Allowance households (DWP, April 2001)
• Adults in Working Families Tax Credit households below a low income threshold (DWP, April 2001)
• Children in Working Families Tax Credit households below a low income threshold (DWP, April 2001)
• Adults in Disability Tax Credit households below a low income threshold (DWP, April 2001)
• Children in Disability Tax Credit households below a low income threshold (DWP, April 2001)

Purpose of Domain

The purpose of this domain is to capture the extent of income deprivation in an area.

Background

Income deprivation is now often measured at national level as the proportion of households below a particular low-income threshold. International comparisons frequently use the proportion of households living below fractions of median or mean income (see Bradbury and Jantti 1999). Thus Eurostat has adopted a definition of income deprivation as those living in households below 60% of median income. National and regional estimates of households below fractions of median/mean income invariably derive from large scale surveys. However, such surveys, even those having a reasonably large sample size do not allow reliable small area estimates. Further data on consumption (and wealth) are collected in a variety of social surveys, but not with sample sizes that would allow reliable small area estimates.

However, despite the lack of comprehensive data on income distribution at a small area level, robust data on means tested social security benefits are available which give valuable insights into low income at very small spatial units. The indicators in this domain are in the form of non-overlapping counts of people living in families in receipt of certain means tested benefits. This domain is presented as the proportion of the population of a ward living in families in receipt of these benefits.

Indicators

Means tested benefits may be divided into 'out of work' benefits, 'in work' benefits and benefits which support housing costs.

'Out of work' benefits comprise Income Based Job Seekers Allowance (JSA(IB)) for those who are unemployed and Income Support (IS) for other groups such as older people, those with a disability or lone parents. Data for April 2001 were obtained from the Department for Work and Pensions (DWP) for these benefits.

'In work' support derives from Working Families Tax Credit (WFTC) and Disability Tax Credit (DTC). WFTC is paid to those in low paid work who have children (both lone and couple parents) as a top up to their earnings. DTC is equivalent to WFTC but for disabled people. Eligibility for WFTC/DTC extends much further up the income distribution than did their predecessors Family Credit/Disability Working Allowance (FC/DWA). Whilst one could argue that all those in receipt of FC/DWA should be counted as income deprived, such an argument is not so easily sustainable for WFTC. However, it is possible to calculate in a modified way whether a particular family has an equivalised income within a particular fraction of national 'benefit unit' equivalised mean/median income. The WFTC data as currently extracted has reliable information on earned income and tax credit in payment. No information is available on housing costs or housing benefits. It was nevertheless possible at this stage to calculate equivalised income for 'benefit units' based on earnings plus WFTC plus Child Benefit, but excluding housing benefit and other income, and presenting the results before housing costs. The DWP Households Below Average Income (HBAI) Unit were approached to run a national profile on HBAI data using both the modified definition of income and on 'benefit units' as distinct from households. In this way they were able to supply a 'cut off' level to enable the inclusion within the domain of the population of those 'benefit units' below 60% median income before housing costs. Ward level data for April 2001 of WFTC and DTC recipients and their dependants below the 60% median income threshold were obtained from the DWP.

For JSA(IB), IS, and those WFTC/DTC cases below the low income threshold, the population (claimant, any partner plus any dependent children) reliant on the benefit were included in the domain and expressed as a percentage of the total population for the area in question.

In general, the in work and out of work benefits do not overlap. There is a very small contingent of IS/JSA-IB recipients who continue to receive WFTC if they become unemployed during the currency of a WFTC award but these account for very few people and can reasonably be ignored.

Combining the indicators

The indicators in this domain were summed in order to generate the percentage of the total population living in such families. The confidence interval of the proportion was such that 'shrinkage' was not necessary in this domain, and the Income Domain score is the unadjusted rate.

Other Issues Considered

Benefit take-up

One of the acknowledged problems of producing a measure of income deprivation using benefits data is that of take-up. The data can easily be adjusted for non take-up provided reliable small area data on take-up are available. Take-up can vary by the type of benefit, the area, the population group (such as pensioners), and over time. The DWP provide data on take-up for different claimant groups at Great Britain level. The latest figures are for 1999-2000 and are not broken down to sub GB level (DWP, 2002). Scotland specific work which is potentially much more useful has also been undertaken (Bramley, Lancaster and Gordon, 2000). Unfortunately this work was carried out on the Scottish House Condition Survey 1996 which pre-dates both the introduction of JSA-IB and WFTC. In sum adjustments for non-take-up could be useful but there are presently the following potential problems:

• Transparency would be lost. Currently this domain represents actual rates of reliance on the benefits in question.
• There are no estimates of take up rates for WFTC - an important component of this domain.
• There are no estimates for Scotland for 2001. The most recent GB estimates are for 2001, the most recent Scottish ones are for 1996.

For these reasons it was decided not to adjust the domain for non take-up, though this issue should be reviewed in possible future revisions.

Benefits relating to Housing Costs

Means tested benefits to support housing costs are Housing Benefit (HB) and Council Tax Benefits (CTB). Most of the recipients of these benefits will be recipients of JSA(IB), IS, WFTC or DTC. There will, however be some who are not, and an investigation was carried out as to whether they could be included in the domain. The Housing Benefit Matching Service of DWP now collects HB/CTB data from almost all local authorities at individual level. These data were obtained from the DWP but after a thorough quality check were found to be not of sufficient quality to include in the domain.

Employment Deprivation

Employment Deprivation: Indicators

• Unemployment claimant count of those aged under 60 (ONS, April 2001)
• Incapacity Benefit recipients aged under 60 (DWP, April 2001)
• Severe Disablement Allowance recipients aged under 60 (DWP, April 2001)
• Compulsory New Deal participants - New Deal for the under 25s and New Deal for 25 + not included in the unemployment claimant count (DWP, April 2001)

Purpose of Domain

This domain seeks to measure enforced exclusion from the world of work. The domain does not seek to capture income deprivation to which joblessness leads, since this is tackled in the Income Deprivation Domain. 'Employment deprived' people are thus defined as those who want to work but are unable to do so through unemployment, sickness or disability.

Background

Conventionally employment deprivation is captured by the monthly claimant count. Whilst this is a good starting point it has become increasingly apparent that it does not tell the whole story.

There has been growing concern that measures based on the unemployed claimant count substantially under-estimate the numbers who would work if work were available. Such groups are referred to as the 'hidden unemployed'. They include those (particularly women) who are seeking work but not registered as unemployed. Some of these people may be captured at Scotland level through the International Labour Organisation (ILO) definition of unemployment contained in the Labour Force Survey. There are also those people on New Deal options who do not appear on the count but who would do so if the New Deal had not been in operation. There are also those people who have taken early retirement. Another group who might be considered are those people who are carers. One of the most significant groups are those people who have moved on to sickness and disability related benefits in the absence of any realistic prospect of finding work (Beatty et al, 2002).

Data to tap into some aspects of 'hidden unemployment', such as those excluded from the claimant count but within the ILO definition, have proved difficult to obtain at the ward level. However, it is possible to count those people incapable of work through sickness and those on New Deal Options.

Indicators

The ONS supplied claimant count data for April 2001 and the DWP supplied data for the same time point (April 2001) for those on New Deal options.

Those who are workless through sickness can be captured by counting those on Incapacity Benefit (IB) and those in receipt of Severe Disablement Allowance (SDA). If the intention is to measure only 'hidden unemployment' then a proportion could be calculated. Otherwise the entire group could be incorporated on the basis that these people all face exclusion from work, whether due to sickness alone or some combination of sickness and labour market conditions. The latter option was selected.

Because men over 60 who are unemployed can choose to receive Income Support rather than Income Based Job Seekers Allowance, ¹ the claimant count for men aged 60-64 is an undercount. Women aged 60-64 are not included in the claimant count. Moreover until the publication of the 2001 Census it is impossible to derive a reliable denominator at ward level which distinguishes women 60-64. For these reasons all indicators in the domain have been restricted to people aged 16-59. Furthermore because the domain is wider than simply those conventionally regarded as 'economically active' the denominator is all persons aged 16-59.

Combining the indicators

As with the Income Domain, the indicators in this Domain constitute non overlapping counts of those excluded from the labour market through unemployment or ill health. A simple rate was therefore constructed - those people aged 16-59 who are unemployed, on a compulsory New Deal (under 25s or 25+), on Incapacity Benefit/Severe Disablement Allowance, are presented as a proportion of all those aged 16-59.

The small size of confidence intervals across the domain did not suggest that the shrinkage technique needed to be applied and the Employment Domain score is the unadjusted rate.

Other Issues Considered

Lone parents

There is a question as to whether lone parents should be incorporated into the Domain. Lone Parents who are not working have traditionally been regarded as 'economically inactive'. They are not required to 'sign on' to get benefit until their youngest child is aged 16. Those claiming benefit do not therefore count as 'unemployed'.

This domain is defined to include those who are involuntarily out of employment. Given the formal position of lone parents, how should those on Income Support be treated? Are they voluntarily or involuntarily out of employment? If the former they have no place in this domain. If the latter they should be counted. This is a sensitive issue. It is impossible to tell whether a particular lone parent on IS has decided that she cannot go to work because her children need her care or whether she cannot go to work because she cannot find an appropriate job or childcare (see Evason et al, 1998). Because the position cannot be known with certainty it was decided not to include Lone Parents on IS in this domain though they are, of course, included in the Income Domain.

Regarding participants in the New Deal for Lone Parents the situation is slightly different. On the face of it, such participation need not signify a wish to re-enter the labour market. However, since April 2000 the initial job-focused interview is now compulsory. Though the full NDLP is still 'voluntary', this compulsory element re-introduces ambiguity - does a participant lone parent really regard herself as involuntary out of the labour market or is she going though NDLP because she sees it as essential to guarantee benefit receipt? On this occasion it was decided to exclude lone parent participants of NDLP from the Employment Domain.

Health Deprivation and Disability

Health Deprivation and Disability: Indicators

• Comparative Mortality Factor (CMFs) for under 75s (ISD, 1997-2001)
• Hospital episodes related to alcohol use (ISD, 1997-2001)
• Hospital episodes related to drug use (ISD, 1997-2001)
• Comparative Illness Factor (CIF) (DWP, 2001)
• Emergency admissions to hospital (ISD, 1997-2001)
• Proportion of population being prescribed drugs for anxiety or depression or psychosis (ISD, 2001)
• Proportion of live singleton births of low birth weight (<2,500g) (ISD, 1997-2001)

Purpose of Domain

This domain identifies areas with relatively high proportions of people who are losing years of life because of premature death or whose quality of life is impaired by poor health.

Background

While ill health is closely intertwined with other aspects of deprivation, it is also an important aspect of deprivation in its own right. It may require unique policy responses and service provision. It is therefore useful to be able to specifically identify geographical areas of health deprivation.

There is a long history of mapping health. However, this work has tended to focus on mortality and certain acute illnesses. There has been far less work carried out on the small area mapping of chronic illnesses, disabilities and health behaviours. The challenge is therefore to update the traditional indicators with new measures. This has largely been achieved through an exploration of various administrative data systems such as prescription, hospital and social security benefit databases.

A number of techniques have been developed to deal with particular problems encountered when constructing indicators of poor health for geographical areas with varying demographies and populated by small numbers of people. These methods include age-sex standardisation and the shrinkage technique. The latter is used for improving the estimate of a rate in an area with a small population.

Indicators

Comparative Mortality Factor for under 75s

When calculating measures of mortality and morbidity for wards, it is necessary to standardise the measures for age and sex to avoid them simply reflecting local demographic profiles. Traditionally indirect standardisation has frequently been used to produce, in the case of mortality measures, a Standardised Mortality Ratio (SMR). However, the literature has consistently pointed out the problems associated with using indirect standardization (see Yule, 1934; Kilpatrick, 1959; Freeman and Holford, 1990; Julious et al, 2001; Sutton et al, 2002). Direct standardization methods have been employed in this Domain to produce a Comparative Mortality Factor. Appendix 3 outlines the arguments and procedures for adopting this approach. This indicator was constructed for people under 75.

Hospital episodes related to alcohol use

Excessive alcohol consumption has both short and long-term health consequences on an individual's mental and physical health. In order to measure alcohol abuse rather than use, acute and psychiatric discharges from Scottish hospitals that could be linked specifically to alcohol were extracted from hospital episode statistics (HES). This included, for example, cases of mental and behavioural disorders due to the use of alcohol, poisoning and exposure to alcohol, and foetal alcohol syndrome. ² It is possible that this measure is affected spatially by service provision and could over or under count all users; however it should not do so for seriously ill users which is the focus of this indicator. The relevant hospital discharges across Scotland were assembled for a 5 year period. The discharges were linked to ward of residence, and a rate was calculated for the estimated resident population in 2000.

Hospital episodes related to drug use

Drug abuse has a significant physical and psychological impact on individuals. It is an especially significant health hazard for younger people, and a major cause of premature death (General Register Office for Scotland, 2001; Jackson, 2002). A count of those admitted to hospital because of their drug use was calculated. Again, it is possible that this measure is affected spatially by service provision, so for example, if drug services are developed differently in different areas, fewer users might then be hospitalised. The advantage of using the hospital data for a measure of the health impact of drug use is that it is known that all those counted are in a poor state of health. Hospital data may lead to an over or under count of all users but not of seriously ill users. The relevant hospital discharges across Scotland were assembled for a 5 year period. The discharges were linked to ward of residence, and a rate was calculated for the estimated resident population in 2000.

Comparative Illness Factor

There are few small area measures of chronic health conditions and yet they affect a large proportion of the population. The census count of long term limiting illness introduced in 1991 represented a major step forward in the measurement of this important aspect of health deprivation. Unfortunately the only indicator available to this project was the 1991 Census which was felt to be too out-of-date to be used. Instead a new measure using health related benefits was created. This combined information on people receiving Disability Living Allowance (DLA), Attendance Allowance (AA), Incapacity Benefit (IB) and Severe Disablement Allowance (SDA).

DLA is a benefit for those severely disabled people under 65 needing help with personal care or with mobility needs. AA is an equivalent cash benefit for people aged 65 or over who need help with personal care but not for mobility needs. People over 65 can receive DLA for mobility needs providing that they were receiving it before they were aged 65. IB is a non-means tested benefit paid to people who are unable to work due to ill health but have paid sufficient National Insurance contributions. SDA is a similar benefit given to people who have paid insufficient contributions to qualify for IB. ³

Because the sets of benefits overlap, it is not possible simply to combine the counts into a single measure. This would lead to individuals being double counted. It would have been possible to calculate two rates for individuals receiving health and disability related benefits: one to measure the proportion of economically active individuals who could not work because of their health, and the other to measure the number of people who received benefits because of their care and mobility needs. However, improvements in the health related social security data available mean that a single non-overlapping count of individuals receiving one or more health related social security benefit can now be calculated. Individual records were linked and those identified as receiving more than one benefit were counted only once.

The counts produced were then age and sex standardised using the direct method outlined in Appendix 3, generating the Comparative Illness Factor (CIF).

Emergency admissions to hospital

Emergency admissions into hospitals capture, amongst other things, two important aspects of health deprivation: externally caused injuries (e.g. accidents or violence) and poorly responding or untreated conditions. The fact that they are emergency admissions is an indicator of their immediate severity.

The measure used was based on a count of those admitted to Scottish hospitals over five years. Discharges, excluding transfers within hospital or to another hospital, from this group were then linked to their resident wards and rates were calculated based on the resident population in 2000.

Proportion of population being prescribed drugs for anxiety or depression or psychosis

Mental health problems affect a large minority of people across Scotland. In some wards over half the population will be experiencing some form of mental ill health. It is therefore an important element of health deprivation to measure. ISD hold a dataset containing a list of drugs that were prescribed by GPs during 2001. From this, all prescriptions relating to anxiety, depression and psychosis were extracted. An average daily quantity (Defined Daily Doses - a World Health Organisation standard) for each drug was used to calculate, from the weight of the total prescriptions, an average count of people being prescribed any one of these drugs. The one year of data was treated as a sample from time. If, for example, a person was given a one week prescription by their GP they should appear 52 times within the year. Each prescription would therefore be counted as 1/52 of a person. By summing the whole year one person would be counted. The method of attributing prescription information to ward level involved linking the individual prescription to a GP practice, and calculating a rate of drug prescription for the practice. This practice rate was then distributed to each person attached to the practice using ISD's practice to patient lookup. The ward level score for this indicator was produced by aggregating the practice rates (now linked to an individual) to a ward, and calculating the mean rate for each ward.

Proportion of live singleton births of low birth weight

Low birth weight (under 2500 grams) is linked to increased morbidity and mortality in infancy. It is also linked to long-term health problems, such as hypertension, coronary heart disease and type II diabetes. Low birth weight is therefore a useful indicator of health deprivation amongst the very young and is also linked to poor maternal health.

The indicator used here combined 5 years of data and was a count of all singleton births that were less than 2500 grams. The denominator was all singleton births. The resulting proportion had the shrinkage technique applied to it.

Combining the Indicators

Combination followed two steps:

All the variables were converted to the standard normal distribution based on their ranks. These new scores were then factor analysed (using the Maximum Likelihood method) deriving weights for their combination.

The variable's ranks were transformed rather than using their raw values, to avoid outliers, possibly resulting from measurement error, having a disproportionate affect on the overall ward scores. The standard normal distribution was chosen as the suitable distribution because there was no 'natural' distribution amongst the variables and because factor analysis (as a parametric based technique) assumes a normal distribution. The first factor explained 57% of the variance and appeared to be a suitable summary measure across all the indicators.

The weights that were derived from the analysis are shown below.

Other Issues Considered

Cancer registration

Cancer registrations were considered as a possible health deprivation indicator. Cancer is not only a major cause of premature death in Scotland, it is also an illness that causes considerable physical and psychological distress to those suffering from it. However because premature deaths caused by cancer would be included in the CMF calculation and individuals who could not work or needed care because of cancer would be counted in the CIF, it was decided that a separate measure was not needed.

Smoking

It can be argued that smoking is an important aspect of health deprivation because of its impact on health - half of long-term smokers will die from conditions related to their habit - losing some 20 to 25 years of life (Bobak et al, 2000). There are no comprehensive datasets on smoking behaviour at a small area level. Any data collected at a primary care level is not gathered centrally and no surveys are large enough to estimate local rates. However women, at the start of their pregnancy, are asked whether they smoke or not. The use of this data was explored for use in the Index. Unfortunately, although this data appeared to present good proxy information on smoking across Scotland, recording errors in a few areas meant that, even had it been considered as an appropriate indicator of health deprivation, it could not be used within the Health Domain.

Education, Skills and Training Deprivation

Education, Skills and Training Deprivation: Indicators

• Working age adults with no qualifications (Labour Force Survey, 1996-2000)
• Pupils aged 16+ who are not in full time education (DWP, 2001) *
• Proportions of the 17+ population who have not successfully applied to Higher Education (UCAS, 1999-2001)*
• Pupil performance on SQA at Stage 4 (SQA, 2001)
• Secondary level absences (Scottish Executive, 2000/1)

* These two indicators were combined

Purpose of Domain

The central purpose of the Education, Skills and Training Domain is to measure in as consistent a way as possible the key educational characteristics of the local area that might contribute to the overall level of deprivation and disadvantage. Many previous attempts to measure educational deprivation at the local level have tended to include both social and educational measures, typically using indicators such as free school meals as a proxy for income deprivation. On the basis of the approach that has been adopted for the Scottish Index of Multiple Deprivation, 'free school meals' is not needed in the Education Domain, as the Income Domain captures children in families receiving Income Support (IS) or Income-Based Job Seekers Allowance (JSA(IB)), which are the eligibility criteria for receiving free school meals. Low income is certainly a correlate and probably, in part, a cause of educational deprivation but it is not a direct measure of educational deprivation as such.

Background

Most measures of educational deprivation have tended to focus predominantly on the school age population. Yet pupils at school represent only one part of the population, which might contribute to the overall educational deprivation of the area. Results from school examinations cover only one particular age cohort, many of whom are likely to move out of the area once they become adults. For this reason the aim of this Domain was to extend the scope to include some measure of the adult population's educational capacity.

Many of the items that were reviewed for the education domain dealt with educational performance, measured by examinations and qualifications. The debate on the meaning of educational disadvantage has increasingly focused on educational results, rather than other possible indirect proxies for educational quality (e.g. pupil teacher ratios). Inevitably the final choice of indicators is in part constrained by what is currently measured and assessed. Though there is some legitimate debate about ways of measuring educational performance, conventional measures of educational attainment and formal qualifications are clearly not only valued within education, but also by the job market and wider society.

It may be argued that to include data on educational performance in this Domain could penalise schools in disadvantaged areas that do well, or conversely reward under-performing areas and their schools. While there may be some unfairness here (effort is not rewarded), the objective fact is that, if - for whatever reason - one area has better educational results than another that may be less disadvantaged in other respects, then this area is less educationally deprived. This has to be correct for the Education Domain. Other forms of economic and social deprivation will be picked up by other domains and measures.

In many cases measures of provision, such as preschool facilities, teacher numbers etc. are themselves influenced by existing allocation policies that may provide extra resources for disadvantaged areas. They are thus likely to be poor indicators of such deprivation.

Indicators

Data for the Education Domain came from a number of different sources. It can be grouped into three types: data from outside the school system, individual level information about pupils which can be directly attributable to the local level, and school level aggregate information that has to be allocated to the local area.

Data from outside the school system

Working age adults with no qualifications

Four separate annual extracts (1996-2000) of the national Labour Force Survey (LFS) at individual level, with a local authority level flag (the so called 'LFSLA'), were combined into a single data set. A regression model was developed to predict the proportion of working age adults with no qualifications. The model was then applied to 1991 Census data at ward level, substituting the LFS variables with the appropriate variables from the 1991 Census. The resulting estimates at ward level were converted to the 1999 ward boundaries and adjusted to fit the local authority level estimates obtained from the LFS. The measure used is the estimate of the proportion of the adult population of working age (25-59) who have no qualifications.

The 2001 Census contains a direct measure of adult qualifications. This could, in due course, replace these ward level LFS derived estimates.

Pupils aged 16+ who are not in full time education

Child Benefit (CB) continues to be paid to carers of pupils in full time non-advanced education (school and Further Education college) up to the age of 18. The declining numbers of young people aged 16+ getting this benefit is thus an indicator of the numbers staying on in full time education above the statutory age in each area. However there are some problems with the data. First, it is difficult to ascertain whether young people aged 16 years in this data set are above or below compulsory school leaving age. The response in other Indices for other parts of the UK has therefore been to focus on those aged 17 and 18. However, a significant proportion of this age group in Scotland will have entered Higher Education. For these reasons it was decided to use the 16+ numbers on Child Benefit as the numerator. The denominator is also derived from the Child Benefit system, but is based on a younger age group as a 'proxy'.

The final indicator used is the negative of the direct measure (i.e. those NOT staying on), on the grounds that this represents a measure of how far pupils in the area persist in education beyond the minimum level. This indicator was merged with the indicator of the proportions of the 17 and over population who have not successfully applied to higher education, before being factor analysed.

Proportions of the 17+ population who have not successfully applied to Higher Education

Data from the University and Colleges Admissions Service (UCAS) covering all applicants to Higher Education from postcodes in Scotland were obtained for three successive years at individual level (1999, 2000, 2001). This data is very well postcoded. A small number of cases were added in from England (as having a home postcode in the relevant country was the criteria - rather than national identity). Only successful applicants were retained (the acceptance rate varies at around approximately 80%). Also, those applying from institutional postcodes were removed. This process in fact removes relatively few cases in Scotland, typically less than 2% of successful applicants in any year. There are between 25,000 (1999) and 27,000 (2001) successful applicants of all ages each year.

UCAS applications cover the full age range. As mature students would be more likely to apply from their own address, the upper age point was set at a point that included the majority of entrants straight from secondary level. Setting the upper age point at 19 years (i.e. under 20) includes 72% of all successful applicants in Scotland. Again the denominator at local ward level has drawn on the Child Benefit data for 2001, and last three year groups of compulsory schooling have been used to match the three years of UCAS data.

The final indicator used is the negative of the direct measure (i.e. those NOT getting into Higher Education). This indicator was merged with the indicator of the pupils aged 16 and over not in full time education, before being factor analysed.

Postcoded School Data

Pupil performance on SQA at Stage 4

Data for the total SQA (Scottish Qualification Agency) results (all age groups) was obtained for the year 2001 in individual format from the SQA, with detailed results and a computed points score using the 'Unified Points Scoring System'. This data was very well postcoded, principally to a home address (or domestic postcode). In the final stages a small number of cases which had identical postcodes to their own school were excluded, as were others with clear institutional addresses. Pupils with postcodes outside Scotland and from overseas (e.g. in the Independent sector) were also excluded.

Results for different Stages were examined to see whether younger or older age cohorts should be included. It was decided to focus on Stage 4, as younger pupils who take SQA assessments at earlier stages appeared to have lower points scores (suggesting that they were only taking some preliminary examinations). Older age groups might have been those adding to or enhancing earlier scores. Using Stage 4 pupils only meant that the SQA results contained some 61,000 cases. All SQA cases also had a school (centre) code.

Data was obtained on individual school rolls and other details for Stage 4 pupils in academic year 2000-2001 for all Scottish secondary schools, including independent and special schools. This data was then matched with the school aggregate information obtained from the SQA data to see how far there were pupils in Stage 4 in these schools who did not appear in the SQA results. This comparison allows some account to be taken of the residual 'non-exam' group. This needs to be taken into account, as it is possible that schools in some areas could submit a smaller proportion of their pupils for SQA. Though it is not possible to identify where these 'non exam' pupils reside, it is possible to weight the results according to the proportion of Stage 4 pupils not appearing in the SQA 2001 results for that particular school.

Special schools: A proportion of pupils from special schools take SQA qualifications. They make up less than 1% of the total in the SQA data. Typically these pupils turn out both to have rather low scores and also, when they are linked to school level data, the proportions in their school in Stage 4 appearing in the SQA data is also often low. After reviewing several options it was decided to retain the results for the pupils in special schools with SQA results, but to take no account of the non-exam pupils in such schools as the effect on a few wards was very substantial.

Independent Schools: a larger group of pupils in the SQA data are from the independent sector (including a few independent special schools). There are 2,600 such pupils in the 2001 data set (4.3%). Schools were excluded which contained no or very few pupils who took the SQA. ⁴ In addition schools that had apparently high take up rates of SQA but very low results (suggesting partial take up of SQA only) were excluded by cutting out those where the school aggregate score was less than the SQA average for all secondary schools (160 points on the unified points score). This procedure cuts out only a small number of additional cases. In the retained cases, pupils in independent schools were also set to a weight of 1. As with special schools, this meant that the pupils not entered for exams were ignored.

The average unified points score for each ward was the indicator used.

School Level Aggregate Information

Secondary level absences

Data was obtained on secondary levels absence details for all maintained Scottish secondary schools for academic year 2000-2001. Independent schools were not included in this data. This data takes the form of 'authorised' and 'unauthorised' absences and is also expressed in terms of the average number of half-day absences per pupil. The data covers S1-S5, but was only available as a school aggregate score. It was decided to use both the authorised and unauthorised absence figures combined, as both in some senses constitute missing education. The method used to 'unbundle' this information to local ward was based on the postcoded SQA data. The school average absences (combining both authorised and unauthorised absences) were allocated pro rata to the wards in which the S4 SQA pupils were located. The ward score is the combined aggregate of these individual pupil values. One issue with this data is that as it is not recorded at all for Independent schools, a few wards are based on very small numbers of pupils. Independent pupils constitute about 4% of pupils overall in the SQA data. A large number of wards have close to zero percent of such pupils, while a small number have more than 35% (and up to 70% in one case). For some of these wards the absentee rate may therefore be based on scores for a minority of pupils, who may be atypical of others in the same ward. However such wards are less likely to be at the deprived end of the distribution.

Combining the Indicators

Shrinkage estimation was applied to all indicators in the Education Domain, with the exception of the adults with no qualifications. This indicator has already been 'shrunk' to a notional local authority score as part of the estimation procedure. ³

The indicators have varying distributions so they were ranked and then transformed to a standard normal distribution.

Factor analysis was then undertaken on the five candidate variables in their 'shrunk' format. All five variables had significant correlations at ward level. The association between the UCAS data and staying on at school was among the highest correlations, with the absentee data having a lower value. As the UCAS data and the staying on rate had different enumerators but the same denominator (Child Benefit) it was decided to combine these into a single variable in the final step of the factor analysis. The argument was that if in areas with small populations this denominator either understated or overstated the correct figure then the combined effect of both variables could have a powerful effect on the final position. It was finally decided to proceed with the five variables reduced to four, by combining the UCAS and staying on rates variables on the grounds that they represent very similar measures and use the same denominators.

The factor analysis indicates a robust single factor solution, with no evidence of a second factor. This single factor explains approximately 60% of the variance and has strong correlation with all the variables in the domain.

Using the results from the factor analysis, the indicators were then combined using the weights derived from the factor analysis. The weights that were derived from the analysis are shown below.

Other Issues Considered

Consideration was restricted to strictly educational variables. As noted, 'resource based' measures were not included, as their distribution might already reflect policies to distribute resources differentially to more disadvantaged areas or schools.

A number of school level variables were also considered. These could in principle have been 'unbundled' to local ward using the distribution of SQA candidates, though they would therefore have to have been restricted to secondary level. However several of these (exclusions, special needs) cover only a small proportion of pupils, and it is difficult to be sure how these should be attributed to local neighbourhoods.

With individual pupil level information being collected across all schools in Scotland in autumn 2002, including a full pupil postcode, there will, in future, be increased scope to use such school based information routinely to measure the characteristics of local areas. This was not available for the present study.

Geographical Access to Services

Geographical Access to Services: Indicators

• Road distance to a GP surgery or health centre (ISD, 2002)
• Road distance to a general stores or supermarket (Market Scan, 2002)
• Road distance to a primary school (Scottish Executive, 2001)
• Road distance to a petrol station (Retail Locations, 2002)
• Road distance to a bank or building society (Retail Locations and Market Scan, 2002)
• Road distance to community internet facilities (Scottish Executive, 2001)

Purpose of Domain

The purpose of this domain is to measure the extent to which people have poor geographical access to key local services. The indicators selected relate to health, food, finance, education, fuel and communication.

Background

Poor geographical access to services is treated here as a component of multiple deprivation as it captures an additional aspect of what it is to be multiply deprived. This domain has also been included in the most recent Indices of Deprivation for England, Wales and Northern Ireland.

The domain measures aspects of access deprivation that are relevant to all people. ⁵ It is important to be able to access key local services in both rural and urban areas.

As a relative newcomer to the measurement of multiple deprivation, this domain could be refined in the future, to take into account the availability of public and private transport in some way, for example. However, even without such refinements it is already a robust and important component of the Scottish Index of Multiple Deprivation. The datasets of the location of services represent the best available at the time (for further details see Scottish Executive, 2002b).

Indicators

Road distance to a GP surgery or health centre

It is essential that people have easy access to a GP surgery or health centre. These provide people with vital primary health care and are often the first ports of call for people with health queries. Over a thousand such sites were used for this indicator.

Road distance to general stores or supermarkets

All households need access to a general store or supermarket for their food and other household provisions. This enables people to sustain a healthy diet. Over three thousand sites were included for this indicator.

Road distance to a primary school

Primary schools are a key service for all children aged 5-8. If children have to travel a long distance to their primary school it adds significantly to the length of their day. Also, though school buses are often provided, it remains difficult for parents to collect their children at other times in the day, for example at the end of an after school club. Over 2000 primary schools were included for this indicator.

Road distance to a petrol station

Access to petrol stations is essential for car owners as a source of fuel, but petrol stations often also sell other commodities such as basic food and health related items (e.g. milk, bread, pain killers). Because of their location, petrol stations can be particularly important for people who live in small rural settlements or in places where there are no alternative outlets that provide similar services. Just under 700 petrol stations were included for this indicator.

Road distance to a bank or building society

Banks and building societies offer a variety of financial services and help people, in general, to manage their financial affairs more effectively. About 1400 banks or building societies were used for this indicator.

Road distance to a community internet facility

Although a luxury commodity some years ago, the internet has become a very useful source of information and services, as well as an increasingly used method of communication. The high cost of PCs prohibits some households from obtaining computers and internet access within the home, making the internet facilities provided for community use particularly important. The Scottish Executive supplied the locations of over 650 community internet facilities within Scotland for this indicator.

Distance Measurement

The distance to the nearest service of each type was measured from the population weighted centroid of each Output Area (OA) in Scotland. ⁶ Distance was measured by road, and was rounded to the nearest 10 metres. The distance was measured on the basis that the shortest travel time is preferable, so motorways, A roads and B roads were prioritised over unclassified roads - in practice for most cases the quickest distance was also the same as the shortest distance.

For islands which do not possess a particular service, the road distance was measured to the island's port; the sea distance from that port to the port of the nearest place that does possess the service was measured, using ferry route data supplied by the Scottish Executive; and then the road distance from this port to the service was measured. This was as refined as it was possible to make the model: in practice though, people on unserviced islands will also be dependent on regular ferries or air travel (and indeed on the weather conditions).

Ward level indicators were created by averaging the distances from each OA centroid within a ward to a particular service.

Combining the Indicators

The indicators were ranked, transformed to a normal distribution and combined using weights which were generated by factor analysis. This was a single factor model, with 68% of the variance explained by the first factor. The weights are as follows:

Other Issues Considered

Indicators held by the Scottish Executive that were not included

The Scottish Executive possess data on a range of other services. The services that were included within the domain were selected by the research team and the Steering Group as key services of relevance to all people within Scotland.

Access to Private Transport

Data on car ownership were considered. However, further analysis would need to be undertaken to gauge the meaning of such an indicator or weight. For example, by treating non ownership of a car as a form of access deprivation, this runs counter to the efforts in many areas to reduce car use by improving public transport provision. Conversely, a recent ONS report found that the proportion of people who claim to have some difficulty in households without a car is nearly twice as great as those with a car (Ruston, 2002).

Access to public transport

Public transport is a key service used by the population at large. Moreover, good access to public transport for everyone reduces demands on the road networks and encourages more environmentally friendly practices. Local buses are the cheapest modes of public transport. Nevertheless, measuring access to local bus services is a complicated task since it involves looking at different aspects of one's journey such as walking distance to the nearest bus stop, total travelling time to destination including changes, frequency of bus services to certain destinations, and so on.

The 1999, 2000 and 2001 Scottish Household Surveys were investigated for information about the transport habits of adults in Scotland, in the hope that an appropriate measure of access to local buses could be calculated. Though the 'travel diary' section contained detailed questions about frequency of buses and distance to bus stops, an indicator (or weight) could not be created for a number of reasons. For example, there were only a very small number of respondents in some wards; a ward level measure of this sort would fail to take into account variation in bus routes and provision (including frequency) within wards; there was not sufficient information to determine an appropriate cut-off point for respondents in a ward saying that they had frequent buses, nor for the weight to be assigned to wards with infrequent buses.

However, a database is in the process of being constructed which will contain the grid references of all bus stops in Scotland. This will be a useful source of information for future attempts to measure access to public transport in Scotland.

Options for Measuring Distance

Alternative ways for measuring distance to services have been explored. For example, the measurement of drive time rather than road distance was considered. The Scottish Executive and SEGIS could have provided information on drive time in the form of travel time zones around services. However this package is limited in the amount of data it can process at any one time. Also, it was decided that it would be best to retain the detail of the actual distance that needed to be travelled.

Non Geographical Barriers to Access

Non geographical barriers to access were considered. Access issues relating to mobility and language were considered as well as the provision of culturally appropriate services. However it was not possible to obtain adequate indicators for inclusion in the Domain.

Other Domains Considered

Housing Deprivation

There is rightly a great concern to measure housing deprivation, to help to inform policy, and to target particular groups of people. There is more than one potential approach to the measurement of housing deprivation, and many possible indicators. One approach might be to focus on the provision and accessibility of housing, while another might be to identify poor quality housing. Indicators might include measures relating to vulnerability in the housing market, access to suitable housing, and the special needs of certain groups to have safe and appropriate housing (children and disabled people for example), as well as housing in need of urgent repair, or in a potentially health-damaging condition. However, up to date data to address these issues were not available at ward level for the whole of Scotland and as a result, no housing domain was produced for the Index. This chapter describes previous approaches to measuring housing deprivation in Scotland, and the data sources investigated for the new Index.

The most recent Scottish Area Deprivation Index was commissioned by the Scottish Office in 1998 and completed by the Department of Urban Studies at Glasgow University. This built on the existing Index of 1995, which was wholly derived from the 1991 Census. The 1998 Index included both census indicators, and more up to date and direct measures of deprivation, and was constructed at postcode sector level. The housing related indicators in the 1998 update were overcrowding (households in permanent buildings who are below the occupancy norm relative to all households in permanent dwellings, 1991 Census); lack of amenities (households in permanent buildings lacking exclusive use of bath/shower/insider WC relative to all households in permanent dwellings, 1991 Census); and vacant dwellings (household spaces classified as vacant accommodation or other, relative to all household spaces, 1991 Census). However, due to the procedure used in constructing the Index, in the final multiple deprivation measure, of the housing related indicators, only the overcrowding measure contributed to the overall score.

With the release of data from the 2001 Census, it will be possible to update the variables used in the 1998 Index. However, although comprehensive, the Census by its nature is not capable of being updated as frequently as might be useful for policy and planning initiatives. Over the next two years housing related neighbourhood statistics will be developed under the guidance of the Neighbourhood Statistics Housing Working Group, part of the Built Environment Statistics Advisory Committee of the Scottish Executive. It is planned that this will result in a substantial number of valuable indicators becoming available at sub- local authority level, such as the number of dwellings in each ward by council tax band, the number of vacant and void houses, housing demand, social rented sector rents, houses in multiple occupancy, tenure, and age of dwelling. These data will help in many areas of planning, regeneration, and social justice. Some of this information will be made available with the release of data from the 2001 Census. However, much of it will need to be collected from local authorites or other local organisations. This means that the projected timetable for the collection of the data does not anticipate completion before 2004.

Clearly, not all of the indicators collected will be measures of 'housing deprivation', although each may contribute to the development of housing deprivation indicators. In addition, even when more data are available, future versions of an Index of Multiple Deprivation will need to address which sorts of housing deprivation are best combined together. For example, it may not be best to combine a measure of poor condition housing with a measure of affordability, as the first indicator is a gauge of the housing stock, whereas the second reflects access to housing, or even the relative desirability of an area. The coherence of the measures will need to be carefully considered.

Data sources considered

Because the data from the 2001 Census will be released in 2003, it was not considered appropriate to incorporate housing indicators from the 1991 Census in the housing deprivation measure. Several other issues were considered in relation to housing in Scotland.

The Scottish House Condition Survey

In the Welsh Index of Multiple Deprivation and the Northern Ireland Multiple Deprivation Measures the housing domain was conceptualised as housing 'stress' and focused on the condition of the housing, rather than the situation of the residents. These domains therefore measured key aspects of poor housing, which may or may not 'map onto' other aspects of deprivation. It was judged that whatever the tenure of the house, or the status of the household, living in housing which is in a poor state was itself a deprivation. By showing that a ward contains housing with a low 'Housing Stress' score, there was no implication that this ward (and individuals within the ward) was not deprived in other ways, or that the ward was privileged.

Analysis was therefore undertaken using the 1996 Scottish House Condition Survey (SHCS) to see if indicators from the survey could be used to form a 'housing stress' domain for the Scottish Index. Before the release of the 2001 Census the survey is the most up to date data that covers the whole of Scotland, and it has relevant information about the condition of housing across all of the housing types and tenures. The survey carried out a household interview and in a large proportion of cases a physical inspection. This analysis followed work for the Northern Ireland Multiple Deprivation Measures's Housing Stress Domain where indicators from the Northern Ireland House Condition Survey were used to provide estimates of poor housing at ward level (see Noble, Smith, Wright et al, 2001; Northern Ireland Housing Executive, 1998).

However, after the results were extensively reviewed by Communities Scotland and the Scottish Executive, the indicators used in the analysis were dropped. None of the indicators was felt to be sufficiently robust to be used as a measure of housing deprivation. Also, the number of sample points in some areas was too low to be confidently used at ward level to measures housing stress. In addition, as the survey is over five years old, significant changes in the housing stock have occurred in parts of Scotland, and this invalidated the results in several areas.

Homelessness

Local authorities in Scotland produce figures for housing applications for their area every quarter. These figures are published and publicly available. However, no sub-local authority level data are available for housing applications. A new initiative has been set up whereby for each application for housing, the postcode of the applicant's last dwelling will be collected. When it becomes available, this information will give a good picture of the areas in which households are experiencing difficulties in securing housing, or where homeless households have come from, at a sub-local authority level.

If data on households in temporary accommodation were available, this might go someway to establishing the extent of homelessness in an area, such as a local authority. In fact, households in temporary accommodation are difficult to record as they may only be in one locality for a short time. In addition, as it is often the case that households find temporary accommodation in areas away from their initial home, at small area level a figure of households in temporary accommodation might be more a reflection of the availability of temporary housing, or local policy.

Affordability

The lack of affordable housing in many areas is an increasing concern. Not only does it restrict the ability of people to live near work or family, but it can also contribute to household overcrowding. However, it is very difficult to compose a standard measure of affordability, as it relates to several factors: the composition of the dwelling stock, the costs of renting or buying a house, and household income. This is also being addressed by the Neighbourhood Statistics Housing Working Group.

Scottish Household Survey

The latest Scottish Household Survey was carried out in 2001 and is therefore relatively up-to-date. It contains variables relating to tenure, household composition, shared amenities and room to person ratios. However, although this would potentially provide information about some aspects of housing deprivation at a local authority level, because of the sample design employed, it is not possible to produce unbiased scores below the LA level. In the absence of ward level indicators, it is not possible to model down any LA scores to the ward level.

Crime and Social Order

Crime and social order are important elements in measuring deprivation at the small area level. Ideally, they would be included in an index of multiple deprivation to help to inform policy and local initiatives. Unfortunately, robust small area data on crime or social order for the whole of Scotland were not available to enable the inclusion of this domain in the Index of Multiple Deprivation. Numerous possible data sets and methodologies were explored. A number of developments are in progress to improve the standardising of crime recording practices, and it is hoped that these advances will enable future updates of the Index to incorporate crime and social order indicators.

Although police data is clearly an important indicator of levels and trends in crime and disorder, other partner agencies also collect a great deal of data relevant to this domain. Ideally, any Crime and Social Order Domain would include data relating to the occurrence of crimes and incidents (i.e. where, when and what type), the offender (who and where) and the victim (who and where). Another valuable input would be data relating to fear of crime and the perception of community disorder.

Police Data

Although advances in the standardised recording of police data are underway, there are still several difficulties to be addressed before the data will be available at small area level for the whole of Scotland. These include questions of geography, recording and access to sensitive information.

Police data on crime is collected at local authority level for the whole of Scotland and published by the Scottish Executive at national level. However, similar data at electoral ward level are not available in a consistent form across the eight Police Forces in Scotland. The aggregation of recorded crimes to geographies other than beat areas is not undertaken regularly and beat boundaries themselves are not normally contiguous with ward boundaries.

Even if police data were available at ward level, there are issues on how best the data could be interpreted. The process of recording a crime - from it being reported by a member of the public or a Police Officer, to it being a recorded crime statistic - is not consistent across Forces. A number of stages are involved in the process, and these stages are not presently standardised across Scotland. For instance, the degree and accuracy of geo-coding of crimes varies across the eight Police Forces, and varies by crime type. Some Forces task one or more officers with manually amending incorrect geographical information. This can involve simply correcting the police beat code to which the crime is allocated. The accuracy of the grid-reference attributed to a case may also be variable.

Investigations were also made into the availability and quality of 'Command and Control' data (a record of each crime and incident reported by a member of the public over the telephone). It was not possible to access this data for the whole of Scotland and there is evidence that the data are not of sufficient quality to incorporate into an index at present. The data suffers from variability in the accuracy and consistency of geographical coding. In addition, it is not representative of actual crime levels. For instance, a Force's 'Command and Control' system may record a car backfiring once as gunshots (if several people report it).

All UK Police Forces are in a state of transition from the fragmented data management structure of the past to the new centrally coordinated National Intelligence Model (NIM). The NIM promotes the sharing of intelligence between Forces to combat not only localised crime, but also cross-border and international crime. The outcome of the NIM, the National Intelligence Database, is expected to be fully operational within 2-3 years. Access to data collected in this way would be of great benefit to future updates of the Index.

Fire Service Data

Data on the occurrence of malicious fires and false call-outs were identified as good indicators of social disorder. Previous research has shown that, when combined together, malicious property fires, malicious vehicle fires, malicious small fires (e.g. rubbish fires) and malicious false call-outs, correlate well with other crime and social order indicators at small-area level. ⁷

Several Scottish Fire Services were contacted and requests submitted for data. However, due to the protracted Fire Brigades Union strike action during 2002, the Fire Services were unable to extract, format and provide the data to the Index Team within the specified time period. It is hoped that future updates of the Index would be able to utilise such data within a Crime and Social Order Domain.

Offender Data

Offender data has been considered as a possible source of indicators for a Crime and Social Order domain. The incorporation of offender data into the domain might be a valuable addition as it locates the offender in terms of their home (or temporary) address rather than the location of the crime itself. Offender data therefore avoids the problem often encountered with recorded crime of certain crime types being concentrated in city/town centres. For example, violent crime (such as wounding) is often alcohol-related and occurs at night in city/town centres. A further example is car crime, which may be concentrated in particular city/town centre car parks or retail centre car parks.

The Scottish Executive publishes annual statistical bulletins containing information on criminal proceedings in Scottish courts. The information relates to the types of crime or offence addressed in court proceedings, sentencing outcomes and the characteristics of convicted offenders. These data are compiled from returns to the Scottish Criminal Record Office (SCRO). Unfortunately, the SCRO data are not designed for statistical purposes, and therefore are not suitable for inclusion in the Index. ⁸

The Scottish Executive also publishes annual statistical bulletins on Social Enquiry Reports, Community Service Orders, Probation Orders and Supervised Attendance Orders in Scotland. These reports contain valuable information on the risk factors associated with re-offending. The results are based on aggregated returns provided to the Scottish Executive by local authorities. However, these data are only collected and released at local authority level and even if they were considered to be suitable indicators, would therefore not be available at ward level for inclusion in the Index.

Victim Data

Another source of information on who is affected by crime or social disorder is data on 'victims'. This data, like offender data, locates crimes in terms of the individuals involved, rather than the place of occurrence. The primary source of information on victims of crime is cross-sectional crime surveys, with the Scottish Crime Survey (SCS) being the largest undertaken in Scotland. The most recent sweep of the SCS for which results are available occurred in the year 2000, with around 5000 participants. This survey investigated not only victimisation, but also included questions on fear of crime, perception of community disorder, attitudes to the Police, and self-reported drug misuse.

Unfortunately, although these data are available at national level, it was not possible to model them down to ward level with any confidence. The Criminal Justice Research Branch of the Scottish Executive is currently in the process of undertaking a Fundamental Review of the SCS, the results of which should be published at the end of February 2003. This Fundamental Review will address issues such as the geographic level to which data can be disaggregated, and it is hoped future data will be released with local authority codes.

Insurance Data

Previous research into the use of home contents insurance premium data as a potential indicator of crime and disorder has revealed a number of weaknesses in this approach (see Noble, Smith, Penhale et al, 2000, p.47). For instance, insurance premiums are affected not just by the predicted likelihood of experiencing criminal victimisation, but also by factors such as local environmental conditions, local economic conditions and local insurance take-up rates. A substantial proportion of households do not have household insurance, and therefore would not be reflected in the premiums. Insurance premium data was not, therefore, deemed to be of sufficient reliability or relevance to be included in the Index.

Physical Environment

Poor physical environment is widely acknowledged to be important, both as a determinant of health, as well as more general well-being. Unfortunately, indicators relating to the impact of the physical environment at small area level, are extremely difficult to construct. Though it was not possible to obtain data and therefore construct a Domain for this Index, the situation regarding data is set to improve. The Environmental Health Surveillance System for Scotland (EHS3) is currently gathering information from a wide variety of sources, including local authorities, the Water Authorities, the Scottish Environment Protection Agency and NHS Board Areas, to interpret environmental data in Scotland. Future versions of the Index will benefit if such data become available in a utilisable form.