Small Area Estimates of Income: Means, Medians and Percentiles - - PowerPoint PPT Presentation

Small Area Estimates of Income: Means, Medians and Percentiles

Alison Whitworth March 2015

1

Small Area Income Estimates and Households in Poverty Estimates

• Model based estimates of average weekly household income for

& the proportion of households in poverty

• Available for middle layer super output areas (MSOAs) in

England and Wales

Four income measure:

1. Total income 2. Net income 3. Net equivalised income before housing costs 4. Net equivalised income after housing costs

Poverty measure:

• % of households with net

equivalised income AHC below 60% of median

Uses and requirements

Used for :

• Policy development and monitoring
• Service planning and marketing strategies

Requirement for detailed estimates at small area levels (LA and MSOA), including:

• Median and mean income
• Income distributions (income of bottom 10% in an

area?)

• Poverty estimates (proportion below 60% of

national median income)

3

Current methods

• Area level models, provide estimates of mean

household income and poverty

• Both poverty and income models use the Family

Resources Survey

• Annual survey of households in UK (estimates are

for England and Wales )

• Census records and administrative data used

for modelling

4

Empirical Best Predictor Method

• Model income at household level to provide

estimates of household income.

1) Model income at household level to obtain estimated fixed and random parameters 2) Generate a number of simulations of household income based on model parameters and estimated conditional distribution by using census instead of FRS covariates to estimate income for unobserved households. 3) Thus you derive median household income for each MSOA by averaging across simulations 4) Estimate error using a parametric bootstrap

5

Initial application of EBP method

• Initial application for North East and South

West regions of England

• Based on 2001 FRS and 2001 census data
• Model uses household level variables,

available in both the FRS and Census.

• Includes MSOA area level variables
• DWP benefits data and council tax
• Application provides proof of concept

6

Median household income estimates by MSOA with approximate confidence intervals.

7

Income distributions

8

Model fit: Quantiles of the residuals

against the theoretical normal quantiles

9

Next steps:

• 1. Application using 2011 FRS and census

data

• For all areas within England and Wales
• Check underlying model assumptions / model fit

investigate other transformations / methods

• Calibrate to direct estimates at region.
• 2. Development of methods so that outputs can

be updated between census years

10