About the area classifications

Area classifications have been produced for Great Britain following the 1971, 1981 and 1991 Census, and produced for the UK following the 2001 and 2011 Census.

The 2011 area classifications cover Output Areas, Super Output Areas (Lower Layer Super Output Areas in England and Wales, Data Zones in Scotland and Super Output Areas in Northern Ireland), local authority districts and health areas.

The 2011 area classifications reflect 2011 Census geographies.

The classification for each geography is separate, with the exception of the health area classification, which is based instead on the supergroups, groups and subgroups produced for the local authority classification. Otherwise, the codes and names for each area classification apply to that classification only.

Where possible we have used different names with the classifications to minimise confusion, but this may not always be possible.

Different statistical techniques were used, as appropriate, for each level of the classifications.

To find out more about the method used in producing each classification, please see the relevant methodology information.

Information is also available on the 2011 Census statistics used in each classification.

With the 2011 area classifications, the clusters are split into five main census dimensions: demographic, household composition, housing, socio-economic and employment. Graphs characterising these areas are available at three levels of hierarchy: supergroups, groups, and subgroups. The exception to this is the 2011 Area Classification for Super Output Areas for which two levels of the hierarchy only were produced (supergroups and groups).

An example of a supergroup within the 2011 Area Classification for Local Authorities is “Urban settlements”. Within this supergroup there are two groups: “Manufacturing traits” and “Suburban traits” and within the first group there are two subgroups: “Industrial and multi-ethnic” and “Urban living”.

Various supporting material is available to support the area classification codes and names assigned for different geographies, this includes cluster summaries, maps, methodology guidance, pen portraits and radial plots.

To measure similarity between authorities, a squared Euclidean distance (SED) measure has been used, which is based solely on the 59 census statistics used in the main classification.

In this context, the range is defined as the range between the two most similar authorities and the two most dissimilar authorities.

Pairs of local authorities are considered to be:

• extremely similar if they have an SED of less than 1% of the range between the two most similar authorities and the two most dissimilar authorities
• very similar if they have an SED of less than 2.5% of the range
• similar if they have an SED of less than 5% of the range
• somewhat similar if they have an SED of less than 10% of the range
• not similar if they are more than 10% of the total range

The Corresponding local authorities file is provided to show, individually, which are the most similar authorities.

With the 2011 Area Classification for Output Areas, an error was found in the labelling of two census variables (numbered 32 and 33 in the supporting documentation) used in helping to define two subgroups. As a result, two subgroups have now been renamed:

4a1 is now Social renting young families (previously Private renting young families)

4a2 is now Private renting new arrivals (previously Social renting new arrivals)

All supporting material was updated on 15 April 2015 to reflect these changes.

Corrections to the 2011 Area Classification for Local Authorities

Introduction

A number of data and methodological issues have been identified since publication of the UK 2011 Area Classification for Local Authorities in July 2015. We have now withdrawn the previously published classification, and all supporting documentation, and reworked the classification to produce a corrected version of the classification.

This note sets out the data and methodological issues that have been identified and corrected with the new version of the classification.

What were the data and methodological issues with the previous classification?

There were seven issues in total, which are summarised in the following section.

1. Errors in the counts of persons employed in the agriculture and manufacturing industries affecting Scottish local authorities

   The errors affected two variables: percentage of employed persons aged between 16 and 74 years who work in the agriculture, forestry or fishing industries; and percentage of employed persons aged between 16 and 74 years who work in the manufacturing industry.

   The errors relate to Scottish local authority data only and to persons employed in two particular industries: agriculture, forestry or fishing, and manufacturing. National Records of Scotland (NRS) discovered an error in the coding specifications for industry. A number of people were wrongly coded to “agriculture” rather than “manufacturing”. The errors were corrected on 6 November 2014 on the NRS website, but these errors and subsequent corrections were not known at the time to the Office for National Statistics (ONS) team responsible for producing the area classification.

   Further investigation indicates that the census data originally obtained from NRS contains 8,360 more people employed in agriculture, forestry and fishing; and 8,360 fewer people in manufacturing and distribution than the corrected Scottish Census data. This affects the local authority calculated percentage of persons aged 16 to 74 years employed in these two industries.

2. Errors in the calculation of the variable “percentage of persons aged between 16 and 74 who are unemployed”

   The errors relate to the incorrect components for unemployed count used in the numerator in the calculation of “percentage unemployed aged 16 to 74 years”. This resulted in double counting for persons aged 16 to 24 years and 50 to 74 years categorised as “unemployed never worked” and “long-term unemployed”, and an under-counting of “unemployed aged 25 to 49 years”.

3. Errors in the calculation of population density in a couple of areas – affecting the variable “number of persons per hectare”

   Due to their small population sizes, Isles of Scilly and City of London were included with Cornwall and Westminster respectively. For the population density variable – number of persons per hectare – the wrong area measurements and population figures were used in the calculation of the population density figures for these two combined areas.

4. Errors in the calculation of the standard illness ratio in a couple of areas – affecting the variable individual’s day-to-day activities limited a lot or a little (standardised illness ratio)

   Due to their small population sizes, Isles of Scilly and City of London were included with Cornwall and Westminster respectively; incorrect standardised illness ratios were calculated for these two combined areas.

5. Errors in the calculation of some of the household variables due to the use of incorrect household denominators in the calculation of household percentage figures

   The denominator has been changed from dwelling type all categories to dwelling type household spaces as the original number of households that supported the classification exceeded 100%. This affects five variables relating to households.

6. A methodological inconsistency when applying the standard clustering technique to group local authorities together with areas with similar characteristics

   The SAS software computer programming code used unintentionally resulted in a K-median rather than a K-means clustering technique, as intended, being applied to the transformed and standardised data variables. K-means is the standard cluster technique.

7. An issue with the number of iterations used to generate clusters of local authorities for each supergroup, group and subgroup

   When the SAS code was originally used to generate the clusters of local authorities, a single iteration of the code was run. However, it is now known that running the code over many iterations may result in an optimal cluster of local authorities for each supergroup, group and subgroup.

None of these seven issues in isolation were thought likely to have much of an impact on the classification itself and in particular, to the way in which local authorities are assigned to individual supergroups, groups and subgroups. However, tests have shown that when these issues are collectively adjusted for, that this does result in changes in the supergroup allocation for a number of local authorities – 88 out of 391 (23%) are allocated to a different supergroup.

We have taken the decision, therefore, to rework entirely the 2011 Area Classification for Local Authorities using largely the same 2011 Census data, but corrected as necessary. All supporting documentation has also been updated and corrected.

How can I distinguish the old and new versions of the classification and are they comparable?

The corrected classification is referred to as the 2011 Area Classification for Local Authorities version 2.

Whilst the new corrected classification uses the same coding convention for supergroups (for example, 1, 2, 3 and so on), groups (for example, 1a, 1b, 1c and so on) and subgroups (for example, 1a1, 1a2, 1a3 and so on), the new codes have a superscript ^r (for revised) alongside the code to distinguish the codes from the codes used in the original classification, for example, 1^r, 1a^r and 1a1^r.

As the classification has been totally reworked, the names of the supergroups, groups and subgroups have changed as well to reflect the characteristics of the local authority areas covered by each cluster. The exception to this is that the supergroup name “London cosmopolitan” has been retained, though the local authority coverage for this supergroup has changed.

Whilst the total number of supergroups (eight) is unchanged, the number of groups has increased from 15 to 16 with the corrected classification, whilst the number of subgroups has reduced from 29 to 24. The reworked corrected classification is therefore not directly comparable with either the July 2015 published version of the classification, or the previous 2001 Area Classification for Local Authorities.