ModelBased Estimates of Average Weekly Household Income, 200102

1
Model-Based Estimates of Average Weekly
Household Income, 2001/02
Shayla Goldring Office for National Statistics
http//www.neighbourhood.statistics.gov.uk
2
Outline

• Need for income information at low geographical
  levels
• What is small area estimation and why it was
  chosen
• Data requirements
• Methodology used to provide estimates and CIs
• Quality Assurance
• Results
• Future

3
Background

• Need for Income data at smallest possible
  geographical level

• Who?
• Central Government
• Local Government
• Private companies
• Academics

• Why?
• Identify areas of
• deprivation
• Locate resources

• Sources of this information could include
• Census question on income
• Benefit data from the DWP
• Estimation Methodology

4
Small Area Estimation Project Methodology

• What is small area estimation?
• Small area estimation encompasses a family of
  techniques used when there are insufficient
  survey data available on the variable of interest
  within each small area to provide a reliable
  estimate.
• ONS surveys are optimised for producing estimates
  at the national and regional level
• The ONS has developed the SAEP methodology in
  order to derive estimates of known precision from
  variables contained in social surveys.

5
Small Area Estimation Project Methodology
FRS Survey 3171 wards sampled
Admin and Census covariates
SAEP modelling mean income
Estimates in 8839 CAS wards in England and Wales
6
Income Types

• Total (Gross) weekly household income
  (unequivalised)
• Net weekly household income (unequivalised)
• Net weekly household income before housing costs
  (equivalised)
• Net weekly household income after housing costs
  (equivalised)
• Family A - 100(unequiv) Family B -
  100(unequiv)

Equiv 100
Equiv 83.33
7
Modelling
8
Estimates

• Obtain estimates using the fixed part of previous
  model

9
and Confidence Intervals

• is derived from the previous model in order to
  determine the confidence intervals

• Once the estimates and confidence intervals are
  obtained they are aggregated using ward
  population estimates to create region estimates

• Family Resources Survey region estimates are then
  divided by the aggregated model-based estimates
  to calculate a ratio. This ratio is then used to
  adjust the estimates and confidence intervals.

10
Covariates

• 28 covariates selected as significantly
  associated with total income, from approx. 150
• These include variables from the following
  sources
• Regional indicators
• 2001 Census
• Council Tax 2001/02
• DWP benefit data
• Land Registry house price data was considered but
  not significant in model

11
Selected covariates
12
Diagnostic Testing

• Check for constant variance in the residuals
• Check for random spread of the residuals

13
External Validation

• England, LAD comparisons with IMD 2004, DCLG
  (2005)

14
External Validation

• England, LAD comparisons with IMD 2004, DCLG
  (2005) London Highlighted

15
Results
16
Results
ONS (2005). House Prices 2001/02. Source
Organisation Land Registry. Office of the
Deputy Prime Minister (2004). Indices of
Deprivation for Wards in England,
2004. Department for Work and Pensions
(2001/02). Income Support Claimants.
17
Future

• Development of methodology to provide estimates
  for SOAs
• Research into the feasibility of providing
  estimates of the distributions of income
• Work is also being carried out within the ONS
  looking into the feasibility of including an
  income question on the next Census

18
QUESTIONS?

• Web link
• E-mail enquiries

www.statistics.gov.uk/methods_quality/downloads/
small_area_est_report/saep1_Prelimsch12_v2.pdf
spatialanalysis_at_ons.gov.uk