Conduct of censuses


Within the United Kingdom the national population is determined on the basis of the national Census. Modern censuses have their origins in western Europe. There are still many countries where there have been no censuses or ones covering only urban populations, or only conducted after a several decades.  Chad has never conducted a census.  The Netherlands and Germany have population registers, where each person is required to register with the local authority when they move house and these countries have abandoned undertaking censuses.

In Great Britain, a census has been carried out every ten years since 1801, except for 1941[1].  It attempts to count all people and households on one day - traditionally where the people spent the night.  The Census is overseen by the Office for National Statistics; formerly it was carried out by the Office of Population, Census and Surveys. Before 2001, to administer the census, areas were organised into Enumeration Districts (EDs) of approximately 200 households.  In 1981 in Scotland EDs were replaced by Output Areas, which use areas covered by postcodes as the building blocks. From 2011 Output Areas were adopted across the United Kingdom.  Output areas have approximately 125 households on average, and a minimum of 40.  The average population in an OA has increased from 297 in 2001 to 309 in 2011.  

Further information on the history of the census is available at:  [accessed 01/10/2018].



A census form is delivered to every household and residential establishment in the country. The forms are completed by members of the household (officially by the 'Head' of the household), referring to the specified date of the census, and returned by post or in 2011, online. Participation is a statutory requirement, and enumerators follow up any households from which no form is returned.  Before 2001 data was requested for people located at a residence on the night of census. From 2001 this changed to those in their usual place of residence, and this was continued in 2011 with a separate section for visitors at each location.  

Face-to-face interviews are carried out with a large sample (over 300,000) of  households, to check coverage and estimate under-enumeration (numbers of households and persons missed by the census).

Data collected

Data is collected on individuals and on households. The exact data set varies from  census to census. Ethnic data was first collected in 1991, and the ethnic group classifications were changed for the 2001 census. These ethnic group classifications remained the same in 2011. Areas currently collected include:

for individuals:

  • demography: age, sex, ethnic group, country of birth, religion, marital status, population mobility. In 2011 the nationality was added.
  • health: general health status, limiting long term disability, provision of unpaid care
  • social class and occupation: economic activity status, occupation, industry.  From these, socio-economic classifications are developed.
  • Education: level of qualifications achieved.

for households:

  • household size and structure
  • number of rooms
  • type of tenure
  • amenities
  • lowest floor level access
  • access to a car or van
  • method of transport to work.

Methods of administration in Scotland and Northern Ireland differ from England and Wales.

How results are analysed

Households are aggregated into postcodes, the key constituent unit. Each residential postcode includes on average about 17 households.  Reference files produced by collaboration between the Royal Mail, Ordnance Survey, and ONS link postcode to geographical coordinates, and to each larger physical and administrative structure (current or past) of which they are a part. Up to 1991, the basic census aggregate unit for analysis was the Enumeration District, the caseload for a single enumerator, but problems were experienced regarding unevenness of size. In 2001 these were supplanted by a new category, the Output Area (OA). These are roughly 125 households, and compact and homogenous as can be derived, where possible following natural and administrative boundaries. These in  turn may be aggregated for analysis and publishing purposes into Super Output Areas (SOA), comprising one or more Output Area. Data is analysed by means of cross-tabulations of census variables at OA level, which can be aggregated to SOA, Electoral Ward, Local Authority level, etc, all the way to the entire country.

Super Output Areas (SOA) are split into two layers; lower super output areas (LSOA) and middle super output areas (MSOA).  Lower super output areas are built from groups of contiguous Output Areas and are typically consistent in population size as across England and Wales (typically four to six Output Areas).  The minimum population for a LSOA is 1000, with a mean of 1500 and there is an allocated LSOA for each postcode in England and Wales.

Middle super output areas (MSOA) are larger in size than a lower super output area and are built from groups of contiguous LSOAs, comprising of a minimum population of 5000 and a mean of 7200.

Link to OA definitions  [accessed 01/10/2018]


How the results are disseminated

Census data and analyses are accessible on-line from the ONS Nomis website ( Small area data is also available on-line from the Neighbourhood Statistics website (  Access may also be provided via some academic websites, though these may only be accessible to a limited range of users. There is also a computer analysis suite, SASPAC, which includes both complete small area statistical data and software for analysis and presentation.   Look-up tables from OAs to higher units, and OA and urban & settlement boundary files for use in geographical information systems are downloadable free of charge from the ONS Open Geographical Portal ( without charge.

Issues of confidentiality

Data supplied on census returns is considered absolutely confidential. No form of analysis or presentation is performed or permitted that would enable any individual to be identified, either directly from census data or when census data is viewed in combination with other available sources of data. ONS operates a strict policy of disclosure protection  ( [accessed 21/10/2015]

that applies to successive unit aggregates. As a general rule,  any cell containing fewer than six persons must either be suppressed or combined with another small cell. This issue can be the factor that determines which geographical level is chosen for release of data.


Health, educational, transport and housing planning.
The denominator for health and other population statistics.
Analyses of population trends on a wide range of areas: for example, health, illness. Resource Allocation
Describing deprivation:  Townsend, Jarman and Carstairs deprivation scores are all Census based. The index of multiple deprivation (IMD2015) assigns a deprivation score to each super output area (SOA) and local authority in England[2]. The 2015 release is the first to use 2011 Census geograhical boundaries. SOAs are made up of groups of output areas.  IMD2015 uses Census data to estimate population denominators.

IMD is an important topic and concept when considering populations.  Further explanation can be found later on.


It is the most complete source of information about the population because it aims to include everyone. It is therefore very good for understanding numbers in small geographical areas.

Used as the basis for estimating populations between census years.

The results of the census are considered the nearest there can be to a gold standard national population.

Data are collected at one time.


Expensive (2011 Census cost approximately £480 million[3]).

Criticisms of the Census include a tendency to undercount children, young men, homeless people, and members of the armed forces. In 1991, it was estimated that 10% of men in their 20s and 8% of people over 85 were missed.  In 2001, most of the criticisms related to possible undercounting of inner city populations.

Only undertaken every 10 years.

Self reporting - accuracy difficult to assess.  The elderly tend to overstate their age or round to the nearest five years, divorced men tend to report that they are single. 

Current ethnicity categories not added until 2001.

The data can take a long time to be released. Take care when interpreting results, especially at small area level when the data will not be so robust.

Where the ranges of possible responses to similar-looking questions changes between censuses valid comparison is not possible between analyses of successive censuses, but this may not be apparent to the user.

There may be systematic bias in the census process. In 1991, the enumeration omitted all homeless people. In 2001, the count of young men was substantially lower than expected, which may relate to a cohort who wished to be unknown for tax registration reasons. Members of the armed forces may be omitted. Some questions may be intentionally mis-answered: in the mid 20th century there was concern about deliberate misreporting  of age by women, and in 2001 there was a nation-wide campaign (unsuccessful) to get  the fictitious Jedi Knights recognised by the census as a religion.

Link to 2011 census analysis index:

Link to 2011 Census quality and methodology information:



1   For more on the history, visit





                                                                       © M Goodyear & N Malhotra, 2007, M Goodyear 2016 and S Seager 2018