Robert Beatty Northern Ireland Statistics and Research Agency

1
Robert BeattyNorthern Ireland Statistics and
Research Agency

• www.nisra.gov.uk

2
Designing small area geographies for census and
social data

• David Martin
• School of Geography, University of Southampton
• ESRC/JISC 2001 Census Programme

3
Overview

• Background why design a new geography?
• 2001 EW output area design
• Output area characteristics
• Output area applications
• Development towards Super Output Areas and
  Neighbourhood Statistics

4
EW Geography briefing

• Unit postcode mean 15 addresses
• Postcode and census geographies incompatible
• 1991 Census enumeration district (ED) mean 419
  persons
• 2001 Census output area (OA) mean 297 persons
• NB England/Wales, Scotland, NI differences
  transferability comments

5
Why design a new geography?

• Overtaken by population change
• Unsuitability of enumeration geography for data
  output
• Population range
• Sub-threshold populations
• No linkage to postal geography/EDs unique to
  census
• Boundary placement/internal variation

6
Case for redesign

• Demands from census users
• Everyday and statutory geographies
• Uniformity of population sizes (all above
  threshold)
• Control over shape (observe settlement pattern
  and topographic features)
• Internal homogeneity of population
• Compatibility with previous census geographies!

7
Output area design overview

• Synthetic building block polygons
• Unit postcodes nesting within wards (Dec 2002)
  and parishes, incorporating topographic features
• Automated zoning procedure
• Iteratively recombination of building blocks
  seeking best trade-off of design constraints
  (Openshaw, 1977)

8
Address-based Thiessen polygons

• Thiessen polygons around individual
  ADDRESS-POINTS intersected with ED, ward, parish
  boundaries and road centrelines

9
Unit postcode building blocks

• Address polygon boundaries dissolved to form
  unit postcode polygon building blocks

10
Transferability issues (1)

• Unit postcodes as building blocks because they
  represent everyday geography
• Could be street blocks or any other small unit to
  which social data can be referenced
• Building from existing digital datasets but not
  a trivial task!

11
Automated output area design
Initial Random Aggregation of Building Blocks
Iterative Recombination
Design Constraints (Contiguity,
Thresholds, Shape, Size, Homogeneity)
2001 Output Areas
12
OA design (1)
Initial random aggregation of postcodes into
potential output areas
13
OA design (2)
Choose one postcode at random as candidate for
swapping into a different output area
14
OA design (3)
Make the swap and evaluate the impact on the
overall solution
15
OA design (4)
If swap does not result in an improvement, go
back to the previous configuration
16
OA design (5)
Choose another postcode at random as candidate
for swapping into another output area
17
OA design (6)
If the swap results in an overall improvement,
keep it as part of the solution and examine a new
potential swap
18
Contiguities, thresholds, urban/rural

• Output areas assembled from contiguous postcodes
  (NB treatment of stacked postcodes)
• Output areas above 100 person and 40 household
  thresholds (NB treatment of sub-threshold
  parishes)
• Initial postcode classification to urban/rural
  based on DETR boundaries

19
Size and shape

• Output areas should be as uniformly sized as
  possible target 125 households
• minimize S(OApop-target)2
• Output areas should be as compact as possible
• minimize distance to OA centroid

20
Distance to OA centroid
PC mean centroid OA centroid
21
Intra-area correlations

• Maximize intra-area correlations (IAC) ratio of
  area level to individual level variance
• Higher correlations greater internal
  homogeneity
• Tenure (4) and dwelling type (7) categories used

22
Combination of constraints

• All constraint statistics recomputed at each
  iteration
• Must always meet contiguity and threshold
  requirements urban/rural if possible above
  threshold
• Population, shape and homogeneity constraints
  combined with equal weighting

23
2001 Output Areas (n175,434)
England and Wales
24
OA sizes summarized
25
1991 Ward 1991 ED Code-Point
Portswood, Southampton
26
2001 OA 1991 Ward Code-Point
Portswood, Southampton
27
Portswood, Southampton
28
2001 OA 1991 Parish 1991 ED
South Molton, Devon
29
County District Ward Output Area
30
Neighbourhood statistics

• In support of national strategy for neighbourhood
  renewal
• Aggregated administrative source records -
  updated
• Durable core geography

31
Developments towards Super Output Areas

• OAs as building blocks Three-tier geography
• Tier one (pop mean 1500 n35000)
• Automated zone design, nesting within census
  wards now created
• Tier two (pop mean 7500 n7000)
• Automated design, not constrained to wards, test
  areas consultation
• Tier three (pop mean 25k)
• Methodology under consideration

32
Greenwich lower layer SOAs
33
Conclusion designing small area geographies

• Recognition that collection geographies are
  rarely the best output geographies
• User demand for small area statistics related to
  everyday use (postcodes, addresses)
• Benefits of stable geographical base
• Knock-on implications for official statistics

34
ESRC/JISC 2001 Census Programme
http//census.ac.uk