Methods
Selection of study areas
The project team worked with the Office for National Statistics (ONS) using mid-year population estimates at various geographical levels to evaluate the potential magnitude and geographical distribution of changes in population size since the 2001 Census. The results of this analysis are reported in Cockings et al, 2009. Using these results, six study areas were selected as being representative of areas with the magnitude and types of change likely to be seen by 2011. The six areas were the Local Authority Districts of Isle of Anglesey, Camden, Lancaster, Liverpool, Manchester and Southampton. Selection criteria included: degree of change (population increase or decrease); type of change e.g. general increase or decrease across the whole area versus change concentrated in one or two specific areas; nested breaches i.e. threshold breaches at different output geography levels; urban/rural characteristics; geodemographic characteristics; areas which are likely to be challenging for automated zone design e.g. coastal; whether the areas were one of the Census Test or Rehearsal areas; and areas for which the research team had local knowledge.
Creation of test data
For each study area, a set of household-level test data which estimated the contemporary distribution of population, tenure and accommodation type for 2007/08 (hereafter referred to as 2007) were created, under secure setting conditions at ONS Titchfield. This was achieved by linking 2001 Census household data with contemporary MasterMap™ Address Layer 2 (AL2) address data and with postcode-level mid-year population estimates provided by ONS. Where AL2 records did not match to a 2001 household, accommodation type and tenure were imputed using geometry/attribute-based rules derived from the characteristics of dwellings from MasterMap™ Topographic layer and the ratios observed between tenure and accommodation type categories in 2001.
Building blocks
Two sets of polygonal building blocks (postcodes and street blocks) were created for each of the study areas. As per the 2001 Output Area Production System (OAPS) methodology used by ONS to create output areas in 2001 (Harfoot et al, 2010), the postcode polygons were created by generating a set of Thiessen polygons around households, intersecting these with roads and railways, and then merging neighbouring polygons with the same postcode to create postcode polygons. A set of street blocks was created using a similar methodology, without the postcode polygon merging stage.
Automated maintenance procedures
The automated zone design software AZTool (developed by Prof David Martin, Samantha Cockings and the GeoData Institute at the University of Southampton) was extended to include functionality for automatically splitting, merging or redesigning existing zones in order to meet user-specified design criteria. The new version of AZTool (written in VB .NET) is much more flexible in terms of the permitted number of targets, thresholds and other design criteria and is able to handle larger numbers of zones.
Evaluation of maintenance procedures
Using population and household thresholds which were likely to be employed by ONS in 2011, output areas which contained too many, or too few, people and/or households were identified. AZTool was then used to automatically maintain (i.e. split or merge) these areas. The implications of employing a top-down (i.e. maintain middle layer super output areas (MSOAs), followed by lower layer super output areas (LSOAs), then output areas (OAs), versus a bottom-up (OA-LSOA-MSOA) approach were evaluated by comparing the statistical and aesthetic qualities of the solutions produced by the different approaches. The implications of using postcodes and street blocks as the basic building blocks for the maintenance were also evaluated by considering their influence on the ability to find solutions, on the statistical qualities of the geographies created and on the ability to link postcoded data with census data.
