ONS Small Area Population Estimates Project

1
ONS Small Area Population Estimates Project

• Andy Bates
• Office for National Statistics

2
Project background

• National Strategy for Neighbourhood Renewal
• led to setting up of Neighbourhood Statistics
• requirement for small area population estimates
• ONS Population Statistics Review September 1998
• found unmet user need for small area population
  estimates

3
Purpose of the project

• To investigate the feasibility of producing
  an authoritative set of small area population
  estimates that would be available on a nationally
  consistent basis (England and Wales)
• which methods and data sources?
• what geography?
• what age/sex breakdown?
• how accurate?
• constrained to ONS LA mid-year estimates

4
Methods - key points

• Potential methods identified and short listed
• Averaging estimates may improve accuracy
• Constraining smaller estimates to larger ones may
  improve accuracy
• Method performance depends on data used and the
  characteristics of an area

5
Short listed methods

• Apportionment
• Additive/ratio change
• Cohort component (ageing on element)
• Combination of above (hybrids)

6
Key postcoded data sources

• Patient registers - Health Authorities
• Births and deaths - Registrars
• Prisoners - Home Office
• Dwellings - Council Tax lists, Postcode Address
  File

7
Key aggregated data sources

• Child benefit - formerly DWP, now Inland Revenue
• Retirement benefits - DWP
• Electorate - Electoral registers
• UK Armed Forces - Defence Analytical Services
  Agency

8
Data sources - key points

• Need to consider data quality
• Datasets ideally need to be available nationally
  and cover sub-populations
• Potential for smoothing datasets over time
• Could borrow strength by combining different
  datasets

9
Administrative data analysis

• Patient Registers (adjusted) 2.9m (5.5) above
  the MYEs
• Electorate (adjusted) 0.9m (2.0) below the MYEs
• Child Benefit (adjusted) 101,000 (1.0) below the
  MYEs
• Older persons 132,000 (1.6) below the MYEs

10
Patient Register comparison with MYEs
11
Patient Register comparison with MYEs
12
Evaluation criteria

• Acceptable
• Adaptable
• Data available
• Good quality estimates
• Robust
• Minimal burden on data suppliers
• Easy to explain/not too complex
• Takes account of population sub-groups
• Timely to produce
• Value for money

13
Evaluation stages

• Data comparison with 2001 derived ward MYEs
• Analyse data according to ward area type - ward
  categorisation
• Create test estimates using different data/method
  combinations
• Evaluate test estimates

14
Ward categorisation

• Purpose - to help evaluate administrative data
  and estimates
• How? By categorising wards according to a number
  of characteristics eg
• small population size
• high proportion of young adults in population
• student areas
• areas with high levels of in and out-migration
• inner urban areas..

15
Ward categorisation
Inner Urban (population density)
16
Ward categorisation
Full-time Students 19-22
17
Ward categorisation
Unemployment
18
Ward categorisation
Non-White Ethnic Groups
19
Ward categorisation
International in-migrants
20
Conclusion

• Research is ongoing
• Recommendation on suitable ward methodology
  March/April 2004
• Investigation to be done for producing estimates
  for other geographies

21
Small Area Estimation

• Jane Longhurst
• Office for National Statistics

22
Outline

• The small area problem
• ONS research and development
• Implementation of methods
• Dissemination of estimates
• Future work

23
Requirements for Small Area Information

• Assessment and allocation of resources
• Policy development and review
• Planning service provision
• General economic and social research

24
The Small Area Problem

• Surveys provide reliable national/regional
  estimates
• BUT For local areas the sample size is small or
  zero
• Administrative/census data is available for all
  local areas
• BUT information not generally appropriate

25
Model-based Estimation

• Model the relationship between survey data and
  administrative/census data for the sampled areas
• Use the model specification with
  census/administrative data to produce estimates
  for all areas

26
Research and Development

• ONS set up Small Area Estimation Project, SAEP
• To research and develop a generic methodology for
  small area estimation from household based
  surveys
• Methodology developed to cater for clustered
  design form of most UK household surveys
• Research project established for estimating
  unemployment levels and rates
• Developed data framework

27
Implementation

• Unemployment
• Labour Force Survey, Claimant Count
• Income
• Family Resources Survey, Census and Admin data
  sources
• Crime
• British Crime Survey, Feasibility study

28
Map showing UA/LADs where LFS direct estimates of
ILO unemployment can be published.
29
ILO Unemployment

• We provide estimates of two measures of interest
  
• total ILO unemployed population
• ILO unemployment rate - proportion of
  economically active population who are unemployed
• 95 confidence intervals
• We provide estimates for
• period 95/96 to 99/00
• UAs and LADs for England, Wales and Scotland

30
Map showing modelled UA/LAD estimates of ILO
unemployment.
31
Income

• We provide estimates of
• average weekly household income
• gross and net - unequivalised
• net before and after housing costs - equivalised
• 95 confidence intervals
• We provide estimates for
• period 98/99
• wards for England, Wales and Scotland

32
Validation

• Internal validation
• External validation
• Methodological validation
• Validation with other data sources
• Validation with users

33
Dissemination

• Publication as experimental statistics
• Estimates, CIs and documentation
• Metadata
• User guidance
• Technical reports

34
Future Plans

• Update ILO estimates with re-weighted LFS data
• Produce income estimates for 01/02
• Extend work on BCS
• Develop methods for other variables for NeSS
• Ongoing research