Identifying consumer archetypes for Ofgem
Moving beyond the ’average consumer’ when analysing the impacts of energy policy
Project duration: March 2012 to August 2012
Update | June 2014
The original (2012) report for Ofgem written as part of this project was updated in 2014 to use the latest available data. Both the original and updated versions can be downloaded using the links at the bottom of the article.
Ofgem – the national regulatory body for the gas and electricity markets – regularly carries out its own analysis of the likely impact of policy proposals on consumers' energy bills.
Their analyses have typically shown the likely impacts on the bills of the 'average consumer' - based on a simple national average (median) estimate of household electricity and gas consumption. Ofgem is aware, though, that there is a substantial amount of variation in energy consumption across different types of households in Great Britain, and that energy policies will impact them differently. So CSE was commissioned to develop a more sophisticated approach to Ofgem's impact assessments.
We have developed an analytical model – DIMPSA (‘Distributional Impacts Model for Policy Scenario Analysis’) – which makes it possible to assess the impact of both the costs and the benefits of policies for domestic energy consumers. Underlying DIMPSA is a comprehensive dataset based on the socio-demographically representative sample of UK households surveyed in the ONS Living Costs and Food Survey (LCF).
We derived household energy consumption values based on survey-reported expenditure on fuel bills and known local fuel costs (by payment method) at the time of the survey.
The resulting DIMPSA dataset therefore includes extensive socio-demographic data about the household, along with modelled annual household energy consumption (for all fuels).
Twelve clusters of 'household types'
Once we had this dataset, we were able to look for groups of different types of energy consumers. We undertook some detailed analysis (using a popular classification method called ‘CHAID*) to identify key socio-demographic variables associated with different household gas and electricity consumption levels. The analysis found 12 clusters, or ‘archetypes’, which describe different types of households with similar fuel consumption levels.
The table below is a summary of the 12 consumer archetypes (using 2014 updated data).
|Archetype||No. of households||%||Mean annual electricity use (kWh)||Mean annual gas (kWh)|
|1) Low-income electrically-heated||909,902||4%||6,130||n/a|
|2) All other electrically-heated||1,717,198||7%||8,912||n/a|
|3) Low-income non-metered fuel-heated||541,473||2%||3,383||n/a|
|4) All other non-metered fuel-heated||1,070,042||4%||4,814||n/a|
|5) Low-income, out-of-work single adults in small 1-bed social rented flats (London)||937,111||4%||2,158||8,495|
|6) Young working adults in rented flats (London)||1,142,581||4%||2,853||10,372|
|7) Low-income single adults (lone parents or elderly) in social rented houses||1,265,857||5%||2,640||10,592|
|8) Younger working families in medium-sized rented houses||2,777,321||11%||3,491||13,595|
|9) “Average” mains gas-heated households||8,242,283||33%||3,585||15,280|
|10) Wealthy working families in 3-4 bed semi’s owned with mortgage||2,331,870||9%||4,588||18,784|
|11) Asset-rich, “empty-nesters” in detached houses in less urban areas||2,580,421||10%||4,098||19,226|
|12) Wealthy working families in larger detached houses in less urban areas||1,602,071||6%||5,306||23,832|
|totals / averages||25,118,129||100%||4,217||15,911|
They are differentiated first by household heating fuel, either mains gas or non-mains gas (electric or non-metered fuel). The non-mains gas subset is further divided into four different archetypes (numbers 1 to 4), and the mains gas heated households (representing over 80% of households in Great Britain) are split into eight archetypes.
Using these 12 archetypes will allow a more sophisticated analysis of the impacts of policies on bills, because:
- Ofgem can now use the average consumption levels of different groups of consumers to underpin quantitative analysis of the impact of policy proposals on their bills, rather than using a broad 'average' which doesn't represent what many households actually experience.
- The archetypes provide simple ‘pen portrait’ descriptions of the types of consumer they represent, which may help in the development of more user-friendly communications.
- The impacts of policies on more vulnerable groups will be more immediately obvious, which should help to ensure they are taken into account when making policy decisions.
Each of the 12 archetypes is a distinct group of households. Each archetype is established because those consumers share some common characteristics. However, it is worth noting that some of those characteristics are not necessarily unique to each group.
The pie-chart below shows the percentage of GB households represented by each archetype (2014 update):
The original full project report, ‘Beyond Average Consumption – Development of a framework for assessing impacts of policy proposals on different consumer groups’, is available here.
The update from 2014 is here.
You can also access some of the key tables from the report, including the summary table describing the Archetypes, in Excel format here.
* CHAID (‘Chi-square Automatic Interaction Detection’) is a popular analytic technique for performing classification or segmentation analysis. It is an exploratory data analysis method used to study the relationship between a dependent variable and a set of predictor variables. CHAID modelling selects a set of predictors and their interactions that optimally predict the variability in the dependent measure. The resulting CHAID model is a classification tree that shows how major ‘types’ formed from the independent variables differentially predict a criterion or dependent variable. CHAID analysis has the advantage that it enables more detailed scrutiny of the socio-demographics of households in each category, whilst maintaining a sufficient number of cases to give reliable estimates of scalar values.