the effect of uncertainty on fuel poverty statistics
DESCRIPTION
The effect of uncertainty on fuel poverty statistics. Laura Williams, Department of Energy and Climate Change GSS Methodology Symposium, 6 th July 2011. What is fuel poverty?. A household is fuel poor if it needs to spend more than 10 per cent - PowerPoint PPT PresentationTRANSCRIPT
The effect of uncertainty on fuel poverty statisticsLaura Williams, Department of Energy and Climate Change
GSS Methodology Symposium, 6th July 2011
What is fuel poverty?
A household is fuel poor if it needs to spend more than 10 per cent
of its income on fuel to maintain an adequate standard of warmth,
i.e. if the fuel poverty ratio > 0.1.
In England 2008:
3.335
million fuel poor households
The fuel poverty model
Uncertainty in the inputs leads to uncertainty in the output…
ModelFuel
poverty estimate
INPUTS OUTPUT
EHS data
Fuel price data
Other data
Uncertainty analysis
This analysis looked at the uncertainty
associated with:
1.Household income
2.Fuel prices
Methodology for estimating uncertainty
To estimate the number of fuel poor households:
1. Using the given data, calculate the fuel poverty ratio for each household.
2. Sum those with a ratio greater than 0.1.
To estimate the impact of uncertainty:
1. Modify the input data according to its distribution representing the uncertainty.
2. Using the modified data, calculate the fuel poverty ratio for each household.
3. Sum those with a ratio greater than 0.1.
4. Repeat many (typically thousands of) times, i.e. a type of Monte Carlo simulation, in order the create a distribution.
UNCERTAINTY IN HOUSEHOLD INCOME
English Housing Survey (EHS)
Householder features
Employment
Income information
Dwelling features
Health
Age
Benefits
Earnings
Insulation
Fuel mix
Savings
Type, e.g. flat
Property age
Composition
Uncertainty in EHS income data
Uncertainty considered for 5 types of income:
1. Earnings
2. Housing benefit
3. All other benefits
4. Savings
5. Other sources (including occupational pensions)
Reasons for uncertainty in the incomes reported:• Respondent may not be fully aware of the income of other
householders and report incorrect information.
• When data are collected in banded amounts (done in order to maximise response rates), e.g. earned income and savings.
Under-reporting is not considered as part of the analysis.
Uncertainty in EHS income data
Information on the absolute uncertainties in reported values of
income from the EHS does not exist.
Used a study of the Family Expenditure Survey (FES) from the late
1990s which compared FES incomes to the National Accounts.
Social security example:Year
FES / NA (% - FES total as percentage of National Accounts total)
1985 98.11986 95.21987 95.51988 93.41989 941990 93.31991 93.11992 96.4
Mean: 94.9 %
Standard deviation: 1.76%Coefficient of variation: 1.85%
Uncertainty in EHS income data
Aspect of incomeCoefficient of
variation
1. Savings 15.9%
2. Earnings from employment 1.6%
3. Housing benefit 8.7%
4. All other benefits 1.9%
5. Other sources 6.8%
Coefficient of variation for each of the 5 income types:
i.e. greater uncertainty associated with reported savings than other income sources
Can then construct
an error distribution
using the coefficient
of variation
UNCERTAINTY IN DOMESTIC FUEL PRICES
Uncertainty in fuel prices
• The main methodology uses mean gas and electricity prices for each region and method of payment combination.
• Gas and electricity price data is sourced from DECC’s Domestic Fuel Inquiry (DFI).
• However, this is a simplification of the real situation where actual fuel prices vary in each region due to different tariffs offered by different suppliers.
• Used supplementary data from the DFI on the spread of fuel prices paid by households across the country to approximate a simple error distributions.
• Examples on the next slide!
Uncertainty in fuel prices
-40%
-30%
-20%
-10%
0%
10%
20%
30%
40%
Emids Ea
st NE
N Tha
mes NW
North
SE SW
S
Wes
t Mids
DD
Min
5%
25%
75%
95%
Max
Variation in domestic bills for direct debit customers, England 2008:
-40%
-30%
-20%
-10%
0%
10%
20%
30%
40%
East
Emids Lond
York
s
Mids NE
Sout
h SE NW SW
Mer
s & N
W
DD
Min
5%
25%
75%
95%
Max
Gas Standard electricity
RESULTS
Results – recap of methodology
To estimate the impact of uncertainty:
1. Modify the input data according to its distribution which represents the uncertainty.
2. Calculate the fuel poverty ratio for each household.
3. Sum those with a ratio greater than 0.1.
4. Repeat many (typically thousands of) times.
Headline results
• The distribution of possible values for number of households in fuel poverty when incorporating the uncertainty in income and fuel prices:
• Mean: approx. 3.343 million households• 95% confidence interval: 3.299 and 3.388 million households
(a range of approximately 88,000 households).
Results – more detailed breakdowns
Total number of households
Estimated number of fuel poor
before uncertainty
Most likely value after addition of uncertainty
Bottom of 95%
confidence interval
Top of 95% confidence
interval
Width of 95%
confidence interval
Width of interval as percentage
of total number
households
Lowest 30% of income 6,502 2,971 2,968 2,929 3,007 77 1.2%
Highest 70% of income 14,906 364 375 354 397 43 0.3%
• Estimates of the effect of combined income and fuel price uncertainty on a variety of demographic and dwelling characteristics.
• Example:
Results – more detailed breakdowns
• Interval for the ‘lowest 30% of income’ group: 77,000 households.
• Interval for the ‘highest 70% of income’ group: 43,000 households.
• The narrower range for higher income households is because these households are less likely to be close to the fuel poverty threshold, and so are more robust to the effects of uncertainty.
Conclusions
• The uncertainty analysis has produced detailed breakdowns of the effect of uncertainty surrounding fuel prices and income on the fuel poverty estimates.
• Statistics for those at the highest risk of being in fuel poverty (e.g. lowest 30% of income) are subject to the greatest uncertainty.
• Many assumptions have been made therefore the results are best viewed as indicative.
A full note on the analysis of uncertainty in the national fuel poverty
estimates is available on the DECC website at:
http://www.decc.gov.uk/assets/decc/Statistics/fuelpoverty/1609-2008-fuel-poverty-uncertainty.pdf
Questions?