SARs NEWSLETTER

SARs NEWSLETTER
No. 3 – July, 1994

ADDING VALUE TO THE SARS

The fact that the Samples of Anonymised Records represent individual-level microdata means that additional summary information can be attached to each case. This provides the possibility of greatly enhancing the dataset. A number of such variables are currently being added to the SARs and are described below.

The SAR household file also provides an opportunity to create a number of relatively complex summary variables which can be used to characterise either the household or individual members of the household; for example, the age of the youngest dependent child in the household, or the number of fulltime earners. These are available to all purchasers of the dataset at no extra charge.

CHANGING YOUR JOB?If you have been using the SARs whilst employed by a non-academic organisation and change your job, you may not continue to use the SARs at your new workplace unless it, too, holds a commercial licence to use the data.

LEAVING THE UNIVERSITY?If you work in an academic institution, please make sure that you notify us, and your CMU rep, if you are leaving. Academic users of the SARs are registered as members of the institution in which they work or study. If you leave that institution you must notify both CMU and the CMU contact person at your site. Your registration will be cancelled from the time you leave. If you are moving to another university, then you must re-register as a member of that institution.

YOU CANNOT CARRY YOUR REGISTRATION
FROM ONE INSTITUTION TO ANOTHER

YOU CANNOT CONTINUE TO USE THE SARs AFTER LEAVING
YOUR REGISTERED INSTITUTION

In this issue of the SARs Newsletter:

ADDING NEW VARIABLES TO THE SARs

A lifestages variable on the household SAR.
This follows the same algorithm as that used to derive the lifestages variable in the Small Area Statistics and LBS. The variable is being added to the SAR files held at Manchester Computing Centre and the algorithm is available on disk or over the network. The frequencies for the variable are given below.

Household Composition Type:
This variable is the main classification of household composition used in the 100% tables (Household Composition [100%] Volume, for example). It consists of 22 basic categories, based on the age, sex and number of adults in a household and the number of dependent children also present.

Household Dependant Type:
This classification was introduced by OPCS for the 1991 Census. Households are defined in terms of the number and age of dependents resident in a household. Dependants are defined as dependent children or a person who has both a limiting long-term illness and whose economic position is either ‘permanently sick' or 'retired'.

Cambridge Occupational Scale Scores
The Cambridge occupational scale is a continuous measure, based on occupation and employment status, that provides an altemative to social class. It is designed to measure social advantage and is based on the assumption that there is social interaction between those with similar life-styles. The scale was derived using multi-dimensional scaling methods using data on friendships and marriages between members of different occupational groups. It therefore provides a finely graded hierarchy, rather than a structure of discrete and homogenous classes. Tests using the scale show that it is at least as good in terms of explanatory power as any existing class scheme in a variety of areas, including social interaction, voting and social mobility. Scale values range from 0 to 85, with a separate scale for men and women in recognition of the fact that the same occupation may have a different social standing when held by a woman rather than a man.

A manual explaining the derivation of the scale in further detail, and providing the score value for each SOC unit group is available from the University of Cambridge: K. Prandy (I992) Cambridge Scale Scores for CASOC Groupings, Working Paper, No. 11, Social and Political Sciences, Cambridge.

The scale scores have been added to both the Individual and the Household SAR for Great Britain. Descriptive statistics relating to the scores are contained in the revised User Guide and are also supplied with the disk version of the scores.

We would like to thank Ken Prandy for providing the scores in a special version for each of the two files.

Goldthorpe Classes
The Goldthorpe Class Scheme has also been added to the Household SAR.

Adding earnings data to the SARs
Discussions are underway with the Employment Department to provide information from the 1991 New Earnings Survey for the occupational groups available in the SARs. It is planned to obtain the median hourly earnings for men and women by full or part-time working and age-group by occupational group. It may also be possible to include some regional differentiation for the figures.

Although this information will not tell us what the individual member of the SAR actually earned, it will provide a good idea of the level of earnings that one might expect that person to be able to command. It will not be possible to provide earnings information for the self-employed, as they are not included in the NES. It is also likely that NES hourly eamings for part-time workers will be rather higher than the true figures as the NES omits from the sample employees who do not pay National Insurance contributions - mainly those earning below the Lower Earnings Limit. Despite these limitations, the data should be of considerable value in estimating the income, not just of individuals, but also (in the household SAR) of families and households.

Population weights
For many applications SAR users will want to weight the data back to the original population, correcting as far as possible for Census under-enumeration and the absence of wholly imputed households. To facilitate this, a set of weights have been calculated for the individual SAR (in the first instance) which are based on the mid-year 1991 Registrar General's population estimates for each SAR area and for individual year of age and sex. However, because the mid-year estimates take students at their term-time address, the weights have been calculated to re-create the census population base of usual residents.

Obtaining derived variables
As soon as variables are available, they are added to the on-line access SAR files held at MCC. They can also be distributed on floppy disk with case numbers attached to allow them to be merged into an existing database.

Availability
The following variables are available now:

Lifestage Household file
Cambridge Score Household and Individual file
Goldthorpe Classes Household file (at present)
Population Weights Individual file
Household Composition Household file
Household Dependent Household file

ADDING VALUE TO THE SAR
Frequencies for the Lifestages variable 1% Household SAR

Aged 16-24 %
With no children aged 0-15 in household
1. In a couple 2.4
2. Not in a couple 8.3
With children aged 0-15 in household
3. In a couple 1.3
4. Not in a couple 4.0
Aged 25-34
With no children aged 0-15 in household
5. In a couple 4.2
6. Not in a couple 4.8
With children aged 0-4 in household
7. In a couple 6.0
8. Not in a couple 1.2
With youngest child in household aged 5-10
9. In a couple 1.8
10. Not in a couple 0.7
With youngest child in household aged 11-15
11. In a couple 0.2
12. Not in a couple 0.2
Aged 35-54
With no children aged 0-15 in household
13. In a couple
14. Not in a couple 11.3
With children aged 0-4 in household
15. In a couple 2.5
16. Not in a couple 0.9
With youngest child in household aged 5-10
17. In a couple 3.9
18. Not in a couple 1.7
With youngest child in household aged 11-15
19. In a couple 2.3
20. Not in a couple 3.1
Aged 55 up to pensionable age
Working or retired
21. In a couple 3.7
22. Not in a couple 3.0
Unemployed or economically inactive
23. In a couple 1.5
24. Not in a couple 1.5
Pensionable age to 74
25. In a couple 8.2
26. Not in a couple 6.5
Aged 75 or over
27. In a couple 3.0
28. Not in a couple 5.1
Base = individuals aged 16+ 422,699

 

SAMPLING PROCEDURES AND SAMPLING ERRORS
Ed Fieldhouse

The sampling procedure used in selecting the individual and household samples are complex and, whilst approximating a simple random sample, cannot always be assumed to behave as one. The sampling procedure is described in the User Guide (section 1.4). Because of the effects of clustering and stratification, sampling errors may be larger or smaller than those expected under simple random sampling. Where sampling errors are being applied to estimates, they should be modified by multiplying the simple random sampling error by the appropriate design factor. The design factor provides a measure of the extent to which the standard error of the complex design differs from a simple random sample. Design factors of greater than one imply sampling errors greater than expected under simple random sampling. Preliminary estimations of design factors have been made using two different methods, the first using sampling point information (for the household file); the second comparing differences between expected and observed errors (for the individual file). Work on this is still underway and estimates of design factors may be subject to modification. Full information is provided in Campbell et al Sampling variance and design factors in the Samples of Anonymised Records to be published later this year.

Household File
For some individual variables in the household file, clustering within households results in sampling errors larger than expected for a simple random sample (ethnic group, country of birth, migrants, qualifications, social class, age etc.). The largest effects are for ethnic group (Table 1).

Most other individual level variables have values near unity. In contrast household variables (including those relating to particular individuals) are subject to stratification effects and all design factors are slightly less than one.

Individual File
Estimates have been calculated based on a comparison of the difference between the SARs and 100% Census data (having subtracted residents in wholly imputed households) across the 278 SAR areas and the sampling errors which would be expected from simple random sampling. The method is described in more detail in CMU Occasional Paper 2.

Individual and household variables on the individual file are less likely to be subject to clustering and may benefit from stratification. Most design factors deviate very little from unity, many being less than one. Again the largest design factors are for ethnic group, though the effects are much smaller than on the household file (Table 2).

Clearly, researchers interested in individual level variables should, where possible, use the individual level file because of its greater representativeness on these variables. Conversely, those interested in household variables will find the household file not only more flexible but probably more representative.

Table 1

Table 2
Estimated Design Factors: Ethnic Group of individual, Household File Example Estimated Design Factors: Ethnic Group, Individual File
White 1.84
Black Caribbean 1.60
Black African 1.83
Black Other 1.51
Indian 1.99
Pakistani 2.27
Bangladeshi 2.37
Chinese 1.87
Other-Asian 1.83
Other-Other 1.60
White 1.15

Black Caribbean

 1.00

Black African

 1.06

Black Other

1.04
Indian 1.26
Pakistani 1.20
Bangladeshi 1.04

Chinese

 1.19
Other-Asian 1.02
Other-Other 1.30

THE SARs ARE NOW AVAILABLE ON CD-ROM
The ESRC Data Archive are making available copies of the data files and accompanying documentation on CD-ROM. The cost for academic users will be £25, which covers the medium. For non-academic users, there will be a £75 charge as well as the standard costs for the data.

Anyone interested in obtaining the SARs in this format should contact CMU.

DERIVED VARIABLES IN THE SARs: FIRST RELEASE
Sue Heath
Since the last newsletter was published, an additional 31 variables have been added to the 1% Household SAR and a further 5 to the 2% Individual SAR- The new variables within the Household SAR fall into three broad categories:

(i) Derived variables at the household level: 22 household level variables which include 15 variables summarising the number of residents in the household with particular characteristics (for example, the number of dependent children, the number of persons with limiting long-term illness; the number of unemployed persons), 4 variables giving the age of the youngest and oldest dependent child and dependant within a household, and 3 variables which indicate whether the household is an all-pensioner household, all-adult household or all student household during term time.

(ii) Derived variables relating to the head of household.- There are 4 variables in this category: economic position, age, sex and social class of head of household.

(iii) Derived variables relating to individuals: This category consists of five aggregations of existing variables: the 378 individual SAR occupational codes have been aggregated to SOC Sub-Minor, SOC SubMajor and SOC Major groups; the SAR Industry codes have been grouped into the SIC Divisions; and subject of higher qualification has been grouped into broad disciplines.

The five new variables in the Individual SAR all relate to individuals: SOC Sub-Major and SOC Major codes; SIC Divisions; broad discipline of subject of qualification; and county codes.

All of these variables can be accessed by on-line users of the SARs. A disk containing the codebook and glossary files for the new variables, along with the SPSS code used to produce them, has been sent to all CMU reps in academic institutions and to all commercial users. Additional copies, including hard copies if required, can be obtained from the CMU.

Equivalent variables at the level of the family will be available in the next few weeks.

Sue Heath (New) joined the Census Microdata Unit in March 1993 as Research Officer. During her time with us she has produced two versions of the User Guide (the second of which is to appear shortly) as well as continuing with her own research. On Ist August, she leaves us to take up a post as a Lecturer in the Department Of Sociology, University of Manchester. We wish her every success in her new appointment and hope that we will still be able to draw on her SARs expertise, should the need arise!

CHOICE OF POPULATION BASES AND UNITS OF ANALYSIS
Ed Fieldhouse
Unlike the SAS and LBS, population bases and units of analysis are not predetermined. This means that users must be careful to select the appropriate population base and unit of analysis before any analysis is carried out. The options are different for the individual and household SAR.

(i) Individual file.
User may choose between the following population bases and should select on the variables RESIDSTA and CESTSTAT as appropriate.
In households:
- Present residents
- Absent residents
- Visitors
In communal establishments:
- Residents (non staff)
- Visitors
- Residents (staff)
Because visitors, by definition, may be absent residents at another address, there is a risk of double counting some people. Normally the user will wish to select either residents or visitors. Students, for example, are instructed on the Census form to record themselves as visitors if enumerated at their term time address. However, they should also be recorded as absent residents at their parental address.

User should also be careful to take note of the population bases for particular variables. For example socio-economic group and social class include people who have been in paid work in the previous ten years. To achieve comparability with other data sources (e.g. the LBS) it may be necessary to select only residents currently in employment.

(ii) Household File.
The household file does not contain communal establishments, but otherwise the same guidelines apply. In addition, however, users should be careful to select the correct unit of analysis from:
- Individuals
- Families
- Households
Normally a table extracted from the household file will give counts of persons in households. Depending on the software used, to obtain data relating to the household or to the family, the user may need either to select one person from each household or family as appropriate.

DO STUDENTS READ THE SMALL PRINT?
Sue Heath
An article elsewhere in this newsletter has highlighted the need to be vigilant in selecting population bases for analysis. In most cases, users will probably want to include present and absent residents, and exclude visitors. However, a note of caution needs to be sounded concerning the resident status of students. Census Definitions (OPCS, 1992) notes that for most persons the answer to the question on usual address is straightforward, but for people who live at more than one address throughout the year, it may prove to be more complicated. The Census form contained a specific instruction that, for students who live away from home during term-time, their home address should be taken as their usual address: thus, if they were enumerated at their term-time address, they should have described themselves as visitors and given their home address as their usual address, whilst whoever was completing the form at their home address should have included them on the form as absent or present residents according to their whereabouts on Census night.

However, close examination of the resident status, term-time address and relationship to household head of students aged 19 or over in the household SAR (Table 1) suggests that the picture is far more complex than might have been anticipated. (The age threshold has been used in order to exclude the bulk of students – typically at FE and sixth form college - who are most likely to be living in the parental home.) By cross-tabulating these variables it is possible to build up a picture of the living arrangements of students and where they consider their 'home' to be - or, alternatively, the extent to which students ignored the small print!

Overall, 45.5% of students were enumerated as either present or absent residents at their term-time address, implying that they considered their term-time address to be their 'permanent' or usual residence. Almost a fifth of students appear to be living with their parents during term-time, whilst 12.1% were themselves enumerated as the head of the household at their term-time address, suggesting that every student in this group happened to be the first or only person entered on the census form who was usually resident at that address. Those students categorised as boarders/lodgers were presumably regarded as usual residents by their landlords and landladies.

A fifth of students were enumerated as visitors to their term-time address, even though they were living there on Census night, implying that the instructions in the census form were accurately followed by the form-filler, with a student's home address being put down as their 'usual address'. The largest number of students within this category were boarders or lodgers of the household head, with the form-filler clearly not regarding their student lodgers as permanent residents. A marginally smaller number were unrelated to the head of household, possibly members of all-student households, whilst 4.6% were themselves the head of an all-visitor household at their term-time address. A small number of students in this group were visitors in the parental home, implying that these student only lived with their parents during term-time and had more permanent residences elsewhere.

Almost a third of students were enumerated as residents at an address other than their term-time address, almost all of whom were categorised as the son or daughter of the household head. Table 1 conceals the fact that three quarters of these students were actually absent from that address on Census night, which suggests that these students were either still at their term-time address or on holiday, but that their parents correctly interpreted the census instructions on usual address and counted their student offspring as residents at the parental home.

By far the smallest group is made up of students who were enumerated as visitors at an address other than their term-time address. Again, the largest group was made up of students enumerated in the parental home. These students may well have been genuine visitors, with permanent residences elsewhere, or their parents may have misinterpreted the instructions on usual address. Alternatively, they may have thought of their children as visitors because they were away from home for the bulk of the year.

What implications does this have for the inclusion of students in SARs-based analysis? Restricting analyses to resident students will include the greatest number of students (76% of all students aged 19 or over, and 98.1% of students aged 16-18), but users should be aware that this population base will capture students in households from which many are absent for most of the year. If one is interested in exploring households as they are constituted for the largest part of the year - i.e., during term-time - then users are advised to include in analyses only those students whose address of enumeration corresponds with their term-time address, regardless of their resident status. This will cover fewer students, but for many purposes will be a more appropriate base to use (see Ed Fieldhouse's article on population bases for details on the possibility of double counting).

Table 1:

TERM-TIME ADDRESS, RESIDENT STATUS AND RELATIONSHIIP TO HEAD OF HOUSEHOLD:
ALL STUDENTS AGED 19 YEARS OR OVER

Residents at their term-time address 45.9% (n=4365)
Relationship to household head: Son/daughter 19.8
Head of household 12.1
Partner 5.4
Unrelated 3.9
Boarder/lodger 3.1
Other 1.6
Visitors at their term-time address 19.5% (n=1854)
Relationship to household head: Boarder/lodger 7.3
Unrelated 6.5
Head of household 4.6
Son/daughter 0.7
Other 0.4
Residents at non-term-time address 30.1% (n=2866)
Relationship to household head: Son/daughter 29.6
Other 0.5
Visitors at non-term-time address 4.5% (n=424)
Relationship to household head: Son/daughter 2.5
Unrelated 1.1
Other 0.9

TOTAL 100.0%

 (n = 9509)

Source. I% Household SAR Crown Copyright.

 

HANDY HOUSEHOLD HINTS...
Sue Heath

A recent visit to OPCS's offices in Titchfield proved extremely useful for providing answers to a number of outstanding queries concerning the 1% Household SAR. Other users may find these points equally useful and/or interesting...
· Where a household contains visitors only, by a quirk of processing visitors were given a family number in common, regardless of whether they did in fact form a family unit. Users will notice, for example, that individuals in some all-student households have been allocated a family number of 1 even though they are clearly unrelated. This does not, however, then feed into the coding of family type: even though they may be allocated a family unit, visitors in the SARs never have a family type. It should also be noted that all-visitor households within the SARs are routinely allocated a head of household; in standard Census output all-visitor households are not analysed by head of household.
· Where the first person on a Census form 'qualified' as head of household (that is, they were aged 16 or over and usually resident), yet was unrelated to the rest of the household who happened to form a family unit, that family will have been rendered 'invisible' by routine processing, as they would most likely have been described as 'unrelated' to household head. The family may only have been identified if the form was manually edited for other reasons and it was noted, for example, that the rest of the household shared a common surname.
· Where the first person on the form did not qualify as head of household, the next person on the form who did qualify was assigned head of household, and the form was then manually edited in order to reassign (where possible) relationships between household members.
· The major exception to this rule relates to the 22 households included within the Household SAR which are headed by a dependent child. In some cases, these households contain adults, but because the resident status of the adults is 'visitor', they do not qualify as head of household. The editing procedure followed by OPCS for the coding of 10% items did not routinely throw out these households as queries for manual editing, but automatically assigned head of household status to the oldest resident child, even though that child may have been aged 15 or less.
· Ambiguous relationships within households were also checked manually. Consider the following example: a household contains three generations - a married couple, with their two daughters, one of whom has a husband and child of her own. If the daughter's husband was (correctly) described as 'son-in-law' and the child as 'grandchild of head', there would be uncertainty as to which daughter the husband and child were 'attached': this example would therefore have generated a manual edit and the form would then be scrutinised for clues as to which family unit the partner and child should be allocated. If there was no additional information to be gained from the form, then family units were constructed on the basis of the order in which people were included on the form. However, the form filler may have spelt out the relationships more explicitly, i.e. 'husband of eldest daughter', rather than 'son-in-law' on its own.

UPDATE OF SARs USER GUIDE
Version 2 of the SARs User Guide will be available in July. A copy will be sent to all CMU contact persons and to all commercial purchasers of the SARs. Users are free to photocopy the Guide, or can obtain further copies from the CMU for £5. As well as containing fuller information on the Northern Ireland SARs and further guidance on sampling errors, the new edition incorporates a series of notes on using the SARs with SAS, SPS S, Quanvert and USAR, as well as guidance on accessing the SARs on the new Cray machine at Manchester Computing Centre.

TECHNICAL HELP
Now that more people are starting to use the SARs, there are likely to be technical queries about accessing the data in different software packages. The SPSS and SAS examples given in the User Guide are held on-line at MCC on the Cray CS6400 machine and can be accessed by all registered users.

NATIONAL ON-LINE ACCESS TO THE SAMPLES OF ANONYMISED RECORDS
Keith Cole
The Samples of Anonymised Records have now been transferred to the Cray CS6400. The SPSS, SIR, SAS and Quanvert versions of the SARs are now available to registered users. In addition, the data can be accessed using the SARs exploration package USAR which has been developed at Leeds University as part of the ESRC/JISC 1991 Census Programme. The versions of the SARs held on-line by MIDAS (Manchester Information, Datasets and Associated Services) reflect version 2 of the SARs and are continually updated by the Census Microdata Unit to contain the most recently available derived variables.

Printed documentation on how to obtain online access to the SARs via MIDAS is being prepared and will be automatically sent to all registered users. On-line documentation is available via the gopher system on the CS6400. When connected to the gopher select 'National Datasets Service', then 'UK Census of population Statistics', and then 'Samples of Anonymised Records'. The complete codebooks and glossary files can be downloaded via this route.

NORTHERN IRELAND
Two samples of anonymised records have been extracted from the Northern Ireland Census. These resemble the GB files in form and content there is a 2% sample of individuals and a 1% sample of households and the individuals enumerated in the house-holds. These data have been purchased on behalf of the academic community by the ESRC, JISC and DENI and will be distributed by the Census Microdata Unit.

At the time of writing, licensing agreements are being finalised with the ESRC and the data files are being documented, labelled and read into a variety of different packages for export to users. The licencing agreements will be similar to those in operation for the SARs for Great Britain. Each university (or non-academic purchaser) will have to sign an End User Licence Agreement and individuals within the institution will have to sign an Individual Registration Document. A Codebook and Glossary file for Northern Ireland will be sent to each registering institution and further copies will be available from the Census Microdata Unit.

During the autumn the Census Microdata Unit will be generating harmonised UK files that achieve the maximum possible comparability across the UK. These will be available IN ADDITION to the files for Great Britain and Northern Ireland.

THE 1ST ANNUAL CATHIE MARSH MEMORIAL SEMINAR

Quota versus Probability Sampling

Tuesday, 8th November 1994, 5.00 pm drinks for 5.30 pm start
London House, Mecklenburgh Square, London WC1

Chair:
Roger Jowell
Speakers include:
Bob Butcher
Peter Lynn
Nick Moon
Colm O’Muircheartaigh

Joint Royal Statistical Society/Social Research Association

NO CHARGE FOR ATTENDANCE

FINAL CORRECTIONS TO THE SARs

OPCS have now supplied new versions of the SARs that deal with all the outstanding data problems. The new files provide the following corrections:

QUALEVEL: Level of qualification
The earlier versions of the data provided no information for visitors. This has now been remedied.

QUALSUB: Subject of highest qualification
For people with three of more qualifications the subject coded was the first, not the highest. This has now been corrected to the highest qualification.

The following variables on the 2% Individual file were derived using the 100% census variables, whereas the rest of the SAR variables were drawn from the 10% sample. This gave some inconsistencies as there was some further editing done on the 10% sample. These variables have now been rederived using the 10% variables.

EARNERS: number of earners in the household
PENSHH: number of pensioners
AGEYDEPCH: age of youngest dependent child

These corrected datafiles will be available on-line from MCC as version 3.0 of the SAR files. New versions of the data will not be automatically sent to institutions, but will be available on request.

THE USE OF THE INDIVIDUAL SAR FOR THE ANALYSIS
OF POPULATION SUBGROUPS IN HEALTH AREAS

Peter J. Aspinall

South East Institute of Public Health
United Medical and Dental Schools
of Guy's and St. Thomas's Hospitals

Introduction
The recent release by OPCS of samples of anonymised records (SARs) provides public health researchers with a unique opportunity to study the characteristics of subgroups of their populations. This development takes place against a background of important advances in health services planning including an increasingly explicit and open debate about the process of deciding priorities at all levels in the NHS. Public health medicine departments in particular have been explicitly encouraged to achieve maximum health gain for their resident populations within available resources through epidemiological and comparative approaches to needs assessment. Interest has also focussed on issues of equity and access for particular population subgroups, as demonstrated, for example, by the establishment by the Department of Health of an Ethnic Health Unit. This article looks at the scope for using the individual SAR in health care research at the population sub-group level, emphasising the appropriateness of the individual SAR area for analysis at a time when radical rationalisation in the service has produced a trend towards larger health authorities and commissioning clusters. It seeks to identify the usefulness of the SAR for those working with socioeconomic status and social structural models to account for health inequalities.

Spatial units of analysis
The last three years have witnessed throughout England and Wales the merging of district health authorities into substantially larger authorities or commissioning clusters, the new agencies in some cases having integrated the commissioning role of the former constituent health and family health services authorities through unified commissioning executives. Taking only those authorities whose status has been defined by Statutory lnstruments, 195 health authorities were in existence in 1991, 187 in 1992,154 in 1993, and just 112 in mid-1994. In the South Thames (East) Region, for example, the 15 district health authorities at the time of the 1991 Census have now become six substantially larger (with one exception) authorities or commissioning agencies. Camberwell, Lewisham and North Southwark, and West Lambeth District Health Authorities have merged to form the Lambeth, Southwark, and Lewisham Health Commission with a combined 1991 Census population of 694,358 residents. Similarly, Brighton, Eastbourne, and Hastings District Health Authorities have joined to form East Sussex Health Authority with a Census population of 690,447. Below these 'primary level' health areas many of the new authorities have defined 'localities' for the purposes of commissioning, often based on individual local authority areas, a trend which may be strengthened by the move towards unitary development authorities in local government.

The Individual (2%) SAR - comprising individual Inner and Outer London Boroughs and large (or combinations of smaller) local authority districts exceeding 120,000 persons - provides an ideal building block for population subgroup analysis within these new health authorities and commissioning agencies. A coterminous boundary between the individual SAR geography and the new health authorities and agencies in the South Thames (East) Region exists for four of the six new areas, producing samples from the individual SAR of over 13,000 persons in, for example, both Lambeth, Southwark and Lewisham Health Commission and East Sussex Health Authority. Samples of this size - substantially greater than can be achieved in local and regional health and lifestyle surveys such as HealthQuest SouthEast (SETRHA 1993) - enable research to be undertaken on specific subgroups of the population such as ethnic groups, while the 19 individual SAR areas in the Region provide information for studies of the population in individual or groups of 'localities' in those parts of the Region where they are local authority defined areas.

The socioeconomic position of ethnic groups in Lambeth, Southwark, and Lewisham Health Commission (LSLHC). A programme of research is being carried out by the South East Institute of Public Health (SEIPH) on the socioeconomic position and health status of ethnic groups in South East London, funded by LSLHC and the (former) South East Thames Regional Health Authority. One of the difficulties of undertaking such research has been the paucity of information on ethnicity at the sub-regional level. Large-scale government surveys such as the General Household Survey and Labour Force Survey have provided information in depth at a national scale in the absence, until 1991, of a census question on ethnic group. In the 1981 Census, for example, such information had to be inferred from the country of birth of the head of each household, producing less reliable data than had been provided from the 1971 Census question on parents' birthplace. The situation was radically transformed in the 1991 Census. Information in the Local Base and Small Area Statistics (and now the Special Migration Statistics) provide a wealth of information on the size, age, and gender structure of ethnic groups and on such characteristics as their household composition, housing characteristics, and economic position.

For those wishing to explore in more detail at the sub-regional level the socioeconomic position of these groups, the individual SAR opens up substantial possibilities. The use of epidemiological standardisation procedures based on national or regional counts in, for example, the Limiting Long-term Illness and Ethnic Group and Country of birth Census Topic Reports and the expensive (and lengthy) process of commissioning special tables can now be complemented by the immediate access to over forty person and household variables in the individual SAR. Work at SEIPH is focussing on the use of the individual SAR to identify aspects of the household and family composition of individuals in the White and Black groups - comprising 74.33% and 18.68%, respectively, of LSLHC's population - and their levels of deprivation and disadvantage. Information is available in the SAR for a total of 13,142 persons in LSLHC, including 9,825 in the White group, 1,293 persons in the Black Caribbean group, and 747 individuals in the Black African group. Analysis of these records (and selected characteristics of the individuals' households and families) has facilitated an investigation of such topics as the age of entry into marriage, the prevalence of different marital states, family composition, household structure, the gender of household heads, and the number of earners in households.

For copies of Figures please contact ccsr

Lone parent families
One particular topic of interest in these investigations has been the position of one-parent families with dependent children because of their prevalence in the LSLHC area. In addition, these families have an adverse socioeconomic position (as measured by Census indicators) compared with couple families with dependent children which in turn has implications for the work of public health medicine physicians. An initial point of access to this issue was an analysis of OPCS's Primary Birth Record for residents in LSLHC. This source provided information on 12,491 births during 1992, including mother's country of birth and whether the child was born within/outside marriage. It revealed that for UK born mothers 46.5% of all births took place within marriage, 14.9% were solely registered, 26.3% were jointly registered by mother and father usually resident at the same address at the time of registration, and 12.3% at different addresses. By contrast, amongst mothers born in the Caribbean, 42.4% of births were within marriage, 19.2% were solely registered, 21.6% were jointly registered at the same address, and 16.9% were jointly registered at different addresses. The figures were different yet again for mothers born in 'Other Africa' countries; a significantly higher proportion of births, 56.3%, were within marriage, 18.8% solely registered, 15.7% registered at the same address, and 9.2% at different addresses. Births to mothers born overseas, however, is a poor measure of births to all mothers in the different ethnic groups, especially amongst the earlier immigrants such as the Black Caribbean group, where UK-born mothers play a significant part in the total fertility. For example, 56% of all residents of Black Caribbean ethnic origin in LSLHC in 1991 were born in the United Kingdom.

The 1991 Census Local Base Statistics also suggested a high prevalence of 'lone parenthood' amongst the Black groups. Such households comprised 6.0% of all households in the White group in LSLHC, 20.6% of households in the Black Caribbean group, and 18.6% of Black African group households, all figures several percentage points higher than those for Great Britain. However, such figures are based on the 100% processed information on household composition which identifies, for example, households containing 1 adult of any age and 1 or more dependent children (implying no relationship between the adult in the household and the child(ren)). The individual SAR sample based on fully processed records enables information to be extracted on the family composition of the individuals that matches Table 19 in the Ethnic Group and Country of Birth Topic Report ('Families') and on other characteristics like marital status and gender of head of household not available for ethnic groups in the Local Base Statistics.

Fig. 1, based on the individual SAR, reveals for dependent children aged 0-15 important differences between the ethnic groups. Of children aged 0 in the White group (i.e. less than one completed year of age on Census night 1991), 62 were in married couple families, 19 in cohabiting couple families, and 32 in lone parent families; of such children in the Black Caribbean group, only 2 were in married couple families, 4 in cohabiting couple families, and 21 in lone parent families. Fig. 2 shows the prevalence of different marital states for individuals aged 16 and over in the SAR sample. This shows that marriage in the Black Caribbean group is delayed until the early 20s (age 22 in the sample), whereas in the Black African and White groups it commences in the late teens. There is also substantially more marriage taking place in these two groups as compared with the Black Caribbean group, most of whom were single in the reproductive age range in 1991.

The health implications of lone parenthood have not been fully addressed in the literature. However, a number of studies have demonstrated that the household income of such families, particularly single lone mothers, is substantially less than that of married couple families. The 1992 General Household Survey (GHS) for Great Britain, for example, revealed that while only 4% of married couple families with dependent children had a usual gross weekly household income off £100 or less, in the case of lone mothers it was 42% and, for single lone mothers, 54%. Moreover, research has shown that the number of individuals counted as having low incomes has risen sharply since the early 1980s, with lone parents and the unemployed faring worse (Giles and Webb 1993). In addition to their worse position on census socioeconomic indicators, studies have indicated that children of lone parents have the worst mortality record of any social group (Judge and Benzeval 1993; Robinson and Pinch 1987), higher levels of impaired physical growth (Cole and Cole 1992), poorer nutrition of both the children and the mothers (Moynihan, Adamson et.al. 1993; Gibney and Lee 1993), and fewer items of safety equipment (Kendrick 1994) than children in couple families. Other studies suggest a relationship between single parent family background and the commencement of experimentation with smoking, drinking and drug use (Brody and Forehand 1993; Isohanni, Oja, et. al. 1993; Fidler, Mitchell et. al. 1992; and Smith and Nutbeam 1992).

The availability in the individual SAR of a variable on earners (number of residents in employment in the household) provides a surrogate income measure for different types of family within the various ethnic groups. This includes, in particular, information on households with no earners, one of the variables (for all households) included in the Department of the Environment's new measures of deprivation). As table 1 shows, almost half (49%) of the Black Caribbean households containing persons in the SAR sample aged 0-15 had no earners compared with about 32% of White households and almost 40% of Black African households. If one takes persons aged 0-4, almost 62% of Black Caribbean households had no earner compared with just over a third of White households and almost half Black African households. Clearly, these figures are associated with the higher incidence of lone parenthood in the Black Caribbean group.

Table 1: Dependants aged 0-15 in the SAR: Households with 0, 1, or 2+ earners in the White and Black Groups in LSLHC, 1991.

AGE OF SAR SAMPLE

0-15 0-4
No. % No. %

White group
0earners 507 32.2 200 33.7
1 earner 536 34.1 230 38.8

2 or more earners

531 33.7 163 27.5

Black Caribbean group

0 earners

148 49.2 69 61.6

1 earner

87 28.9 25 22.3

2 or more earners

66 21.9 25 22.3
Black African group

0 earners

88 39.5 46 49.5

1 earner

 81 36.3 34 36.6

2 or more earners

54 24.2 13 14.0

Conclusion

In the USA researchers have demonstrated the substantial value of Public Use Samples in studies of population subgroups since their first release in the mid-1960s (see, for example, Longino and Serow 1992; Burr 1990; Longino, Wiseman, Biggar, et. al. 1984; Clifford, Heaton, and Fuguitt 1982; Rives 1981; Longino 1982). As with these samples in the United States, researchers can now use variables such as the relationship to head of household, marital status, age, and occupation on the individual SAR file to reconfigure individual, household and family information to fit differing analytical needs. Smith has suggested, for example, that the availability of such samples now means that 'bureaucratically defined and normatively constrained categories' like household head can be interpreted, modified or reconfigured to suit socially constructed interpretations (Smith 1992). The advent of the SARs, together with the incorporation of 1991 Census data into the Longitudinal Study, now provides researchers with an opportunity to substantially advance our knowledge of the socio-economic position of ethnic groups. For public health medicine physicians and others who argue that the relationship between ethnic group and health is confounded with differences in socioeconomic status such as income, occupation, and educational attainment, the opportunity is now available to develop more sensitive socioeconomic status models. The scope to develop new household and family classifications equally promises new advances in the study of social structural relations within the different ethnic groups.

Acknowledgments

All data presented in this paper are Crown Copyright.

This work is based on the SARs provided through the Census Microdata Unit of the University of Manchester with the support of the ESRC/JISC/DENI.

References

Brody, G.H., Forehand, R. (1993). Prospective associations among family form, family processes, and adolescents' alcohol and drug use. Behaviour Research and Therapy, vol. 31, no. 6, pp. 587-593.

Burr, J.A. (1990). Race/sex comparisons of elderly living arrangements: Factors influencing the institutionalization of the unmarried. Research on aging, vol. 12, no. 4, pp. 507-530.

Clifford, W.B., Heaton, T., Fuguitt, G.V. (1982). Residential mobility and living arrangements among the elderly: Changing patterns in metropolitan and nonmetropolitan areas. International Journal of Aging and Human Development, vol. 14, no. 2, pp. 139-155.

Cole, T.J., Cole, A.J.L. (1992). Bone age, social deprivation, and single parent families. Archives of Disease in Childhood, vol. 67, no. 10, pp. 1281-1285.

Fidler, W., Michell, L. et. al. (1992). Smoking: A special need? British Journal of Addiction, vol. 87, no. 11, pp. 1583-1591.

Gibney, M. I., Lee, P. (1993). Patterns of food and nutrient intake in a suburb of Dublin with chronically high unemployment. Journal of Human Nutrition and Dietetics, vol. 6, no. 1, pp. 13-22.

Giles, C., Webb, S. (1993). A guide to poverty statistics (low income families and households below average income). Fiscal Studies, May, 14(2), pp. 74-97.

Isohanni M., Oja, H., Moilanen, I. et. al. (1993). The relation between teenage smoking and drinking, with special reference to non-standard family background. Scandinavian Journal of Social Medicine, vol. 21, no. 1, pp. 24-30.

Judge, K., Benzeval, M. (1993). Health inequalities: new concerns about the children of single mothers. British Medical Journal, 306: 677-80 (13 March).

Kendrick D. (1994). Children's safety in the home: Parents' possession and perceptions of the importance of safety equipment. Public Health, vol. 108, no. 1, pp. 21-25.

Longino, C.F. (1982). Changing aged metropolitan migration patterns, 1955 to 1960 and 1965 to 1970. Joumals of Gerontology, vol. 37, no. 2, pp. 228-234.

Longino, C.F., Serow, W.J. (1992). Regional differences in the characteristics of elderly return migrants. Journals of Gerontology, vol. 47, no. 1, pp. S38-S43.

Longino, C.F., Wiseman, F-F., Biggar, J.C., Flynn, C.B. (1984). Aged metropolitan-nomnetropolitan migration streams over three census decades. Joumals of Gerontology, vol. 39, no. 6, pp. 721-729.

Moynihan P.J., Adamson A.J., Skinner, R. et. al. (1993). The intake of nutrients by Northumbrian adolescents from one-parent families and from unemployed families. Journal of Human Nutrition and Dietetics, vol. 6, no. 5, pp. 433-441.

Rives, N.W. (1981). Designing census public use samples for aging research. Research on Aging, vol. 3, no. 4, pp. 375-380.

Robinson, D., Pinch, S. (1987). A geographical analysis of the relationship between early childhood death and socioeconomic environment in an English city. Soc. Sci. Med., 25, pp. 9-18.

Smith, C., Nutbeam, D. (1992). Adolescent drug use in Wales. British Journal of Addiction, vol. 87, no. 2, pp. 227-233.

Smith D. S. (1992). 'The meaning of family and household: Change and continuity in the mirror of the American census. Population and Development Review 18, No. 3 (September 1992).

South East Thames Regional Health Authority (1993). Health Quest SouthEast Regional Report, 116pp.

PUBTRAWL: EARLY WARNING

As part of the reporting requirement attached to use of the SARs, we shall be contacting all registered users at the end of August to ask for details of their research and publications. This publications trawl is being coordinated by Phil Rees and will cover use of all the census products. It is essential that you reply to this survey, even if you have not used the data. This will enable us to assess how widely the SARs are being used and will allow a list of publications to be generated that can, in turn, inform other users.
ESRC Contact SARs Support | CCSR
These pages are maintained by the SARs support team.
Send us comments on this web page.