Creating Improved Census Ward Population Estimates


ONS’ GSS policy of best-fitting 2011 Census estimates (and other national statistics) from output areas has been causing issues for some local authorities that have seen ward boundary changes since 2003. SASPAC has received considerable interest from users on this particular issue as well as discussion on the StatsUserNet and Knowledge Hub forums.

The Issue

All 2011 Census ward estimates have been created using a method of best-fitting from the population weighted centroids of the underlying 2011 Census Output Area (OA). The population weighted centroid is a single point within the OA that represent the median location of households. The whole population of this OA is then assigned to the ward in which it falls. This means that where OA boundaries no longer match ward boundaries (i.e. they have changed since 2003 when the 2001 Census OAs were produced on an exact-fit to wards) the population/household estimates may not reflect all households physically located within the ward boundary. Conversely households outside the ward boundary may be included.


2011 best-fit of OAs to Wards: OA4 is allocated to WARD 2 because OA4’s population weighted centroid falls within the WARD 2 boundary (Source: ONS, An overview of best-fitting: Building 2011 Census Estimates from Output Areas)

Creating improved estimates

This is leading many organisations to look out how they can create improved best-fit estimates using alternative methods of redistributing the OA population more accurately. A number of methods can be used such as:

  • using the proportion of each OA’s areal extent that falls within each ward (assumes the population is evenly distributed across the OA) or,
  • using Census day address lists (e.g. Royal Mail Postcode Address File or the Local Land and Property Gazetteers maintained by local authorities) to distribute the population. Tees Valley Unlimited have very helpfully documented their approach here.  

Another approach will be possible when the 2011 Census Unit Postcode Headcount data (Postcode estimates tables are currently expected in May (p35 of the 2011 Census Prospectus (.pdf)) are published. These will provide details of the number of males, females and occupied households for each Census day postcode and will provide a very accurate method of redistributing the population to ward boundaries.

How SASPAC can help?

Once you have created the improved OA population distributions, SASPAC’s rezoning tool can then be used to easily recreate all the Census tables for these more accurate ward household or population distributions. For more information in how to rezone in SASPAC see Module 4 of our training manual.

Census estimates for other areas such as Parishes have also been best-fitted from whole OAs. This re-estimation method can be applied to any zones where best-fitted figures are insufficiently precise.

But remember…

Creating your own local ward estimates will produce alternative figures to those officially produced by ONS and should be documented and correctly labelled as such to avoid any confusion. Also, the best-fit methodology will be applied to create other ONS datasets, so care should be taken to use denominators created using the best-fit method.

Some useful resources:

ONS page on exact and best-fit estimates

ONS overview of the best-fitting methodology

Download median population weighted centroids for output areas, super output areas and workplace zones in shp and mid/mif format.

ONS National Postcode Directory

ONS 2001/2011 Census geography lookups products