DATA: united kingdom: census: basis

Introduction

The following description of the 1991 Census is taken from:
HMSO, 1992, 1991 Census Definitions Great Britain, HMSO, London (ISBN 0 11 691361 4)

Legal basis

The 1991 Census was taken under the authority of the Census Act 1920. The Census Order, directing that a census be taken in Great Britain on 21 April 1991, was made by Counsellors of State on behalf of HM the Queen in Council on 14 February 1990.

Number of questions

The 1991 Census form contained 25 basic questions. This was four more than in 1981. The new questions were on ethnic group, limiting long-term illness and term-time address of students, and on weekly hours worked. This last question had been included in the 1971 Census but omitted in 1981. The amenities question included a new section on central heating which replaced the section in the 1981 question on outside WCs. Furthermore, a method based on identifying and classifying household spaces in multi-occupied buildings with a separate or shared entrance, provided a count of dwellings that was not available from the 1981 Census.

There was no separate question on employment status in the 1991 Census but categories for the self-employed were included in the economic position question. Information on managers and supervisors is obtained from the occupation question.

Full and sample processing

The 1991 Census questions have, like those in the 1981 Census, been divided into those to which the responses are easy to code (most often answered by means of a tick box) and those to which the responses often require write-in answers and are thus more difficult - and consequently more costly - to process. In general, this division determines whether the question is fully processed (a 100 per cent item) or whether it is processed only for a ten per cent sample of household forms and a ten per cent sample of persons returned on forms for communal establishments (a 10 per cent item).

In the 1991 Census the 100 per cent questions were:

sex and date of birth;
whereabouts on Census night;
usual address;
term-time address of students;
usual address one year before census;
country of birth;
ethnic group;
long-term illness;
Welsh/Gaelic language;
economic activity in the week before the Census;
type of accommodation and extent of sharing;
number of rooms;
tenure of household;
household amenities;
availability of cars and vans; and
lowest floor level of accommodation (Scotland only).

The 10 per cent questions were:

relationship in household;
hours worked;
occupation;
name and business of employer (industry);
workplace;
journey to work; and
higher qualifications.

The field force

The 1991 Census was organised in the traditional way, with the delivery of the self-completion forms by enumerators to households and to managers in charge of communal establishments, and their subsequent collection. Most forms were delivered during the period 12-19 April and it was planned to collect the completed forms in most areas between 22 and 25 April; for certain inner city areas this collection period was extended to 28 April, to include the weekend after Census day. In practice, however, it was necessary to extend collection in many other areas to include the weekend, and in some inner city areas, into the following week. This was mainly because of difficulties experienced in contacting some households, and omissions on the forms necessitating call-backs by the enumerators.

Enumeration Districts

The country was divided into around 130,000 small areas (Enumeration Districts) and to each of these areas an enumerator was appointed, though in some cases enumerators were responsible for more than one district. In England and Wales an average enumeration district (ED) comprised about 200 households, containing about 400 persons. In Scotland EDs were, on average, slightly smaller.

ED boundaries were drawn up by the Census Offices in such a way as to produce reasonable workloads for enumerators to complete their tasks of delivery and collection of the forms in the allocated time. The basis of planning EDs was the expected number of households in the area, but allowances were made for other factors which have an effect on the workload. For example, there is a considerable variation between the work of enumerating a scattered population in a rural area, or in a city centre with mixed residential and commercial premises (where even finding the living accommodation is a task in itself), and the work involved in an area of suburban semi-detached housing. Account was also taken of other factors, such as the percentage of persons expected to be living in communal establishments, the extent of multi-occupied housing, and indications that households would not have English as their first language.

EDs were defined so that they generally did not cross the boundaries of administrative areas such as counties, local authority districts or wards. In England and Wales, the 1991 boundaries also took into account, as far as possible, the 1981 Census ED boundaries and the requirements of local authorities for statistics for special areas. In Scotland, as in the 1981 Census, EDs were defined as aggregations of unit postcodes and took account of the boundaries of postcode sectors, districts and Island Areas.

In England and Wales, the ED is, as in 1981, the basic area unit for standard output, such as the Small Area Statistics (SAS). In Scotland, EDs are not used for output; instead, Output Areas (OAs) are created that nest, as far as possible, within the areas used for SAS in 1981.

Confidentiality

Restricted access to data

A census is taken solely to produce statistics, and steps have been taken to safeguard the anonymity of the individual person or household. Confidentiality is protected by the terms of the 1920 Census Act as amended by the Census (Confidentiality) Act 1991, which provides legal penalties for the unlawful disclosure of census information.

Though names and addresses were required to be entered on the Census form, these are only used:

to help the form filler identifiy each person on the form;
to enable the Census Office to refer to the appropriate person if information on the form was missing;
to indicate whether or not the householder had fulfilled his or her legal obligation to provide the required information;
to enable validation checks and follow-up surveys to be directed to specific individuals;
to enable the linking of a sample of records into the ONS Longitudinal Study; and
for other statistical purposes, for example, in identifying types of family.

No information on identified individuals or households from the Census, or from any follow-up surveys or secondary analysis, is revealed to any persons not working for the Registrars General. On completion of the data processing, the Census forms themselves will be kept securely within the Census Offices and treated as confidential for 100 years, and only in the year 2092 will they be made available for general inspection by the public. Precautions are also taken to ensure that statistical tabulations neither directly, nor indirectly by linkage with other sources of data, reveal information about identifiable individuals or households.

Thresholds for the release of Small Area Statistics

A small proportion of 1991 EDs in England and Wales have very small populations... In order to lessen the risk of inadvertent disclosure of information about identifiable individuals or households in small EDs, the Small Area Statistics (SAS) will not be released for areas with fewer than 50 usually resident persons and 16 resident households (but a count of persons present, residents and households will be provided). Statistics for such suppressed EDs are merged with those of a contiguous ED, provided that the resulting amalgamated total numbers of persons and households also exceed the same thresholds.

Modification

Counts in the SAS abstracts based on the 100 per cent processing at ward level and below are modified to counter the risk that an individual could be identified through a table and hence information about that individual inadvertantly disclosed. All such counts are modified by the addition of +1, 0 or -1 in quasi-random patterns.