A new definition of fuel poverty in Scotland: review of recent evidence

A report by a panel of independent experts who conducted a review of the definition of fuel poverty in Scotland.

Chapter 4 The UK's technical definitions: Boardman and the LIHC indicator

4.1. Common ground

At present, the two technical definitions used in the UK are those developed from Brenda Boardman's definition (used in Wales, Northern Ireland and Scotland) and by John Hills (used in England). Both agree that fuel poverty should be considered a unique form of poverty, distinct from other types of poverty, and requiring tailored solutions. Both also specify a range of parameters that must be objectively measured and monitored - always in the same way over time - to yield a consistent estimate of fuel poverty prevalence and its broad demography. For both, prevalence is estimated by the Building Research Establishment ( BRE) using:

  • a set of indoor temperatures that homes should be able to maintain; this is known as the satisfactory heating regime;
  • the costs that are associated with maintaining those temperatures;
  • additional costs associated with non-heating energy needs, such as for lighting and appliance use.

Both Boardman and Hills set these energy costs alongside the income of a household. Through this they are able to estimate whether the energy cost of attaining a satisfactory heating regime places an undue burden on people's income. If so, the household is deemed to be in fuel poverty.

Both definitions focus on required fuel costs (referred to in past times as needs to spend), rather than actual energy expenditure. This concept was first introduced in 1991 in the English House Condition Survey's Energy Report. Among households that can be effectively captured through required fuel costs are those who should (for example) be spending more than 10% of their income on heating, light and other energy needs around the home, but are not doing so because of concerns about affordability. That cohort is likely to include many of those most likely to need assistance in coping with cold and damp homes. In England, for example, 80% of households in fuel poverty were found to be underspending on energy, when their actual spend was compared to the cost associated with achieving an acceptable heating regime. This fell to 65% of households who are not fuel poor ( BRE, 2013).

However, the LIHC indicator equivalises energy costs, so that household occupancy is taken into account when estimating how much energy a household requires, whereas the Boardman approach does not. The two definitions also assess income differently. Boardman's definition uses income before housing and other costs have been taken into account, and does not equivalise it; the LIHC indicator uses an after housing cost measure which is then equivalised.

Under the LIHC indicator, information on required fuel cost is sufficient to exclude a household from being fuel poor, as long as the required fuel cost falls below the national median required fuel cost. If a household's required fuel cost is above that median, then a further test is carried out, to assess whether their residual income falls below 60% of the national median income; if so, then it can be classified as fuel poor.

For Boardman on the other hand, it is the ratio of the energy cost: household income which determines whether a household is fuel poor or not. Hence, before eligibility for fuel poverty schemes can be determined, both energy cost and income must be assessed in every case.

4.2. The ratio of income to expenditure versus a floating median

The Boardman-based definition currently in use stipulates that a household is in fuel poverty if its required fuel cost is more than 10% of household income. 10% of income is a fixed proportion, which does not change over time. It has been integral to her definition since 1991. At the time, it approximated twice-median actual spend on energy in the UK relative to gross income. It was also the proportion of income the poorest income deciles were already spending on energy.

As a threshold for what could be considered high energy expenditure, twice-median had been a common construct before Boardman. In 1977, for example, Isherwood and Hancock defined ' households with high fuel expenditure as those spending more than twice the median (i.e. 12%) on fuel, light and power'. The median quoted by Isherwood and Hancock was based on the 1977 Family Expenditure Survey.

The choice of twice median expenditure (rather than average expenditure) reflected endorsement of the concept of relative poverty. Medians are more helpful in representing relative concepts than are averages because they are able to smooth out the effects which extreme scores have on averages.

The UK Fuel Poverty Strategy ( DEFRA, 2001), adopted a ' 10% cut off point', following directly from Boardman:

' The 10% cut off point has been used for many years now. The 1988 Family Expenditure Survey ( FES) showed that households in the lower three income deciles spent, on average, 10% of their income (not including Housing Benefit or ISMI as part of their income) on fuel for all household uses. It was assumed by researchers in the fuel poverty field that this could be taken as representing the amount that low-income households could reasonably be expected to spend on fuel'.

It is unclear why a threshold based on the 1988 Family Expenditure Survey was adopted for the UK's 2001 Strategy, since more recent data on energy expenditure was available at the time. However, the 10% cut-off did approximate twice-median actual spend in England at the time the Strategy was formulated.

This is no longer the case. Twice-median actual expenditure is now lower than this in England, and also in Scotland and Wales (the exception being Northern Ireland), as can be seen on Table 4.1. In Scotland, twice-median (2010 - 2012) was 7.4% of gross income.

Table 4.1.: Actual fuel spend as a proportion of income (2010-2012) (Scottish Office data)

Actual fuel spend England Wales Scotland Northern Ireland UK
% of gross income Mean 6.9 6.2 5.9 9.9 6.8
Median 3.3 4.0 3.7 5.4 3.5
% disposable income Mean 6.6 6.5 6.4 11.1 6.7
Median 3.9 4.5 4.3 6.3 4.0

Moving to needs to spend - rather than actual spend - Table 4.2. provides information on the median ratio of required energy costs: income (the fuel poverty ratio) for Scotland, before- and after-housing costs ( BHC and AHC respectively). The three-year average median ( BHC) is 7.5%, making twice-median required energy cost 15% rather than 10%. However, among households who are neither income poor [10] nor fuel poor (the most ideal category), twice median is 11% rather than 10% BHC (or 12% AHC); it is, perhaps, that subgroup whose energy: income ratio could be considered the better reference point for determining who is defined as fuel poor.

Table 4.2.: Required spend: median fuel poverty ratio (2013 to 2015) ( SHCS) [11]

Fuel and Income Poverty Combined Groups BHC AHC
Fuel poor and income poor 18.1% 22.5%
Fuel poor, not income poor 13.3% 13.7%
Income poor, not fuel poor 8.4% 8.6%
Neither income nor fuel poor 5.5% 5.9%
All Households 7.5% 8.7%
Sample 7998 8024

The LIHC indicator also uses median energy costs to assess whether households are in fuel poverty. By contrast, this is a floating rather than a fixed median, based on national energy costs prevailing in the year that fuel poverty prevalence is being assessed. Households whose required fuel costs are above the current median (which will change each year) are, potentially, in fuel poverty - though only if these costs mean that their residual income leaves them in income poverty

Whether a floating median is any more or any less problematic than a fixed median has been much debated, for example:

' By defining 'reasonable' as 'less than the median', the energy costs threshold produced by Professor Hills becomes relative and arguably arbitrary in nature: half of households would always fall beneath it and be facing 'reasonable' fuel costs (whilst half of households would always be facing 'unreasonable' fuel costs). The shifting nature of the median means that it is difficult to reduce the fuel poor headcount through efficiency improvements; as energy costs reduce, so will the median.' (Moore, 2012b).

The extent to which a floating median evens out annual estimates of fuel poverty prevalence can be seen in Table 4.3 which compares estimates of fuel poverty in Scotland using:

  • Boardman, where a fixed median is applied, and where >10% of income for energy = fuel poor;
  • the LIHC indicator, where a floating median is applied, and where above current median energy costs = fuel poor provided residual income is below the poverty threshold.

Table 4.3.: Percentage of fuel poor households in Scotland 2007 to 2015 - Boardman and LIHC indicator estimates ( SHCS)

2007 2008 2009 2010 2011 2012 2013 2014 2015
Boardman (B) 25% 27% 33% 28% 39% 35% 36% 35% 31%
LIHC 13% 12% 13% 13% 12% 11% 12% 12%
B - LIHC +12% +15% +20% +15% +23% +25% +23% +19%

If fuel prices had risen in line with inflation in Scotland between 2002 and 2015, the rate of fuel poverty in Scotland (2015) would have been 8% rather than 31% (Hansard, 2017), illustrating the extent to which the LIHC's floating median obscures one of the principal drivers of fuel poverty, namely fuel prices.

As is well known, national fuel poverty strategies in Scotland and elsewhere in the UK have been predicated on alleviating, if not eradicating fuel poverty. The LIHC Indicator also makes achieving one of these extremely difficult (alleviation), and the other almost impossible (eradication), since prevalence will always calibrate around the floating median. To address this problem, Hills introduced the concept of a 'fuel poverty gap', which is discussed in the next section.

4.3. Severity of fuel poverty: Boardman and the fuel poverty gap.

In the current era of austerity, there has been growing interest in targeting scarce resources towards those in deepest fuel poverty (Walker, Liddell, McKenzie & Morris, 2014). This makes a metric which can reflect different severity levels of fuel poverty especially important.

Based on the Boardman definition, the Scottish House Condition Survey ( SHCS) can be used to generate a ' 5-fold indicator' of fuel poverty. This is provided in Table 4.4.

The five categories represented on the table are defined as follows:

  • Extreme fuel poverty encompasses households whose required fuel costs are more than 20% of their income;
  • Severe fuel poverty encompasses households whose required fuel costs are over 13% but no more than 20% of their income;
  • In fuel poverty encompasses households whose required fuel costs are over 10% but no more than 13% of their income;
  • Not in fuel poverty encompasses households whose required fuel costs would consume 10% or less of their income. Some of these will be in so-called 'marginal fuel poverty;
  • In marginal fuel poverty encompasses households whose required fuel costs are more than 8% but no more than 10% of their income.

Table 4.4.: Prevalence of fuel poverty: five-fold severity indicator ( SHCS)

5-fold indicator 2010 2011 2012 2013 2014 2015 Mean 2013-2015
Extreme 10% 9% 9% 10% 9% 8% 9%
Severe 13% 13% 13% 14% 14% 11% 13%
Fuel poor 12% 12% 13% 12% 12% 12% 12%
Marginal 12% 12% 14% 13% 13% 12% 13%
Not fuel poor 53% 55% 52% 52% 52% 57% 54%

Between 1990 and 2012, when all 4 nations used the Boardman definition, a more common distinction in annual reports was between fuel poverty (energy costs require more than 10% of income) and a subset of households in extreme fuel poverty (more than 20% of income). Table 4.5 contains details. Two features are noteworthy from the Table:

  • the proportion of fuel poor households who are in extreme fuel poverty has been consistently just above a quarter of all the fuel poor in Scotland since 2010 (27% - 28%);
  • more than 200,000 households were estimated to be in extreme fuel poverty in 2015.

Table 4.5.: Fuel Poverty and Extreme Fuel Poverty in Scotland (N in '000s (%) ( SHCS)

Year Fuel poverty Extreme fuel poverty Extreme as % of all fuel poor
2010 818 (35%) 225 (10%) 28%
2011 779 (39%) 209 (9%) 27%
2012 824 (35%) 222 (9%) 27%
2013 860 (36%) 236 (10%) 27%
2014 845 (35%) 229 (9%) 27%
2015 748 (31%) 203 (8%) 27%
3 year mean 818 (34%) 223 (9%) 27%

The LIHC definition generates a more precise estimate of severity through a measure called the fuel poverty gap (see Table 4.6). Instead of the broad bands of severity the LIHC indicator calculates severity to the nearest £1.

Table 4.6.: Fuel poverty gaps in Scotland ( SHCS data)

Year Median fuel poverty gap in Scotland via LIHC Change in gap
2010 £511
2011 £505 -£6
2012 £520 +£15
2013 £545 +£25
2014 £591 +£46
2015 £532 -£59

Given that the use of a floating median, adjusted each year, greatly constrains annual variations in LIHC fuel poverty prevalence, it is the fuel poverty gap which gives the primary estimate of severity in the LIHC indicator.

For that reason, it is a crucial element of the LIHC metric for:

  • monitoring change in national estimates of how serious or otherwise fuel poverty is;
  • identifying people most in need;
  • assessing whether intervention programmes are making inroads.

4.4. The demography of fuel poverty - Boardman & LIHC compared

The lack of any read across from one definition to the other is further evidenced by differences in the types of households each definition finds to be most vulnerable to fuel poverty (Preston et al., 2014). This means that the overlap in terms of which types of households are most likely to be fuel poor is small.

The same has been found in France, where only one-third (35%) of fuel poor households were common to both approaches. As with studies in the UK, the LIHC indicator detected more low-income households, more families, more tenants and more homeowners with a mortgage.

Most of these differences can be explained by the differences between the two indicators in:

  • use of fixed versus floating thresholds;
  • the deduction of housing costs under the LIHC indicator;
  • the equivalisation of income and energy costs under the LIHC indicator (Imbert, Nogues & Sevenet, 2016).

Under Boardman, a drawback of perhaps greatest magnitude is that the definition does not preclude wealthy people from being classified as fuel poor. Households on a net income of £60,000 would be fuel poor under the Boardman definition if they lived in a large, draughty and uninsulated old home with a required fuel spend greater than £6,000 (even though they may not actually spend that amount).

The same household could not be classed as fuel poor by the LIHC indicator. Although their required fuel costs would be patently higher than the median, these costs would still leave them with a residual income above the poverty threshold (60% of median income).

To most Scottish practitioners, it is the classification of wealthier households as fuel poor which creates the greatest unease with the Boardman definition. It has major impacts on prevalence. In fact, the Boardman definition, as used in Scotland, does not align successfully with section 95 of the Housing (Scotland) Act 2001 which indicated that:

'a person lives in fuel poverty if that person is a member of a household with a low income living in a home which cannot be kept warm at a reasonable cost'. (bold font added)

Between 2013 and 2015 less than half of all fuel poor households in Scotland were also income poor (46%). As can be seen in Figure 4.1. the confluence of households that are both income and fuel poor under Boardman is a relatively small one (estimated at 360,000 households) when taken in the context of the Scottish population of households (15% of all Scottish households).

Figure 4.1.: The population of households (N = 2.43M) by income and fuel poverty status ( SHCS, 2015)

Figure 4.1.: The population of households (N = 2.43M) by income and fuel poverty status (SHCS, 2015)

According to the Scottish Fuel Poverty Evidence Review of 2012, and the SHCS 2015, households who are fuel poor but not income poor can be characterised as follows:

  • they tend to live in rural areas;
  • in detached and energy inefficient homes which are under-occupied;
  • their homes are heated for a significantly longer period (8 hours more per week);
  • the majority of them are in work;
  • they have the highest self-reported spend on energy (almost £1,500 per annum) but also the greatest shortfall from what they should be spending in order to achieve a satisfactory heating regime (a needs to spend estimated as £2,400 per annum);
  • more than three-quarters of them (82%) would be classified as vulnerable under Scotland's definition of the term, mainly as a result of the current age threshold is set at pensionable age;
  • they are marginally less likely than others to feel fuel poor (5% rate themselves as subjectively fuel poor, compared with 7% of the population as a whole).

Under the LIHC indicator, concern is of the opposite kind, centred on who fails the fuel poverty test, rather than who passes it erroneously. Here, a group who are frequently not considered fuel poor are relatively low income households who live in small [12] , reasonably energy efficient homes. These households are likely to fail the first test under LIHC, namely whether they fall below the national median energy cost.

As Moore et al. (2012b) point out:

' The majority of low income households with energy costs below the median have the potential for low cost energy efficiency improvements that would save at least £50 per year - in practice likely to be much higher. Can a low income household's energy costs be considered 'reasonable' if their homes can be made cheaper to run at low expense?'

For England, Moore and colleagues estimate that this omission excludes 12% of what would be construed as fuel poor households under Boardman. For Scotland, Figure 4.2 indicates that the exclusion may involve a greater proportion of households: 15.0% of Scottish households were classified as low income/low cost ( LILC), which was almost 3% more than households classified in fuel poverty ( LIHC) (12.3%).

Figure 4.2.: Households in each quadrant of the fuel poverty indicator ( SHCS 2015).

Figure 4.2.: Households in each quadrant of the fuel poverty indicator (SHCS 2015).

The LILC group merits careful consideration for at least 2 reasons. First, they may be among the most cost-effective households to fuel poverty proof. Second, they could be among the more needy households, in terms of the strain which energy bills place on their meagre disposable income. This possibility has been confirmed using English House Condition Survey data, where LILC households are found to be unable to afford required household energy costs once other essentials (such as rent, food and clothing) are costed against their income (Moore, pers. comm.). As energy prices rise, and the floating median rises along with them, the burden of energy bills on these particular households is likely to escalate, but without any improved prospect of them passing the energy cost above median test.

4.5. Affordable heat or affordable domestic energy?

Both Boardman and LIHC estimate fuel poverty based on all domestic energy needs. This is in line with the original intention of Boardman's work on fuel poverty (1991) which was to include all energy needs (i.e. heat, light, and appliances). [13] If the rationale for tackling fuel poverty centres on public health, then some have argued that fuel poverty should be calculated solely from what is needed to achieve a satisfactory heating regime, omitting the energy required for cooking, lighting, etc. In that vein, an alternative concept, 'heat poverty' has been mooted (Scottish Government, 2012).

The latest available data (from the Scottish House Condition Survey) indicate that the modelled cost of space heating makes up just under half of the average domestic energy bill, so a measure solely based on heat poverty would have a substantial impact on the prevalence of fuel poverty. Based on the 2008-2010 SHCS data, for example, it would have reduced fuel poverty from 28% of households to 11%.

It would require only a simple amendment to either the Boardman or the Hills definition to achieve this modification. However, almost one-third (31%) of the average domestic energy bill derives from cooking, hot water, lights and appliances. It is difficult to imply that these basic energy requirements are of less importance to people's wellbeing and quality of life than are warm rooms.

The work which has been completed in Scotland on Minimum Income Standards (e.g. Hirsch et al., 2016) is relevant here, since 3 of the 7 elements making up these Standards require significant expenditure on energy, and 2 of these implicate electricity:

  • 'housing and domestic fuel' - costs associated with heating a home and hot water are derived from BREDEM 12, and so reflect the 'needs to spend' methodology which both Boardman and Hills have endorsed;
  • 'household goods and services' - these include a range of small electrical goods (lamps, hairdryer, hair straighteners, kettle, toaster, iron, hand blender);
  • 'Social and cultural participation' - TV, laptop and internet access.

Hence, there are reasonable grounds for estimating a combination of heating and all other domestic energy needs when calculating required fuel costs.

4.6. Affordable domestic energy or a healthy indoor environment?

Whilst the issue of safe indoor temperatures is examined in detail later on (see Chapter 5), stakeholders that we consulted on heat-only or heat-plus-electricity options also reminded us of the extent to which a healthy living environment should encompass more than safe temperatures - condensation, ventilation, damp and mould contribute to indoor climate too. As a consequence, the revised definition was one which we thought should be concerned less specifically with "an affordable heating regime" and more broadly with the attainment of a healthy indoor climate [14] .

4.7. Under-occupancy

Under the Scottish definition of fuel poverty, under-occupancy occurs when a house exceeds the bedroom standard of its occupants by 2 or more rooms. Estimates of fuel poverty are made on the assumption that all under-occupied rooms require heating to the same standards as the main bedroom [15] (18 °C), which increases estimates of required fuel costs.

The original rationale for heating all rooms of the house was contained in the Parker Morris Report of 1961:

' Better heating…provides an extra degree of freedom in meeting individual needs in the areas of the home which at present are too cold to be suitable for daytime and evening use except in the summer'.

The Scottish method for estimating fuel poverty prevalence is in keeping with these remarks, although not for the reasons that Parker Morris proposed. Instead, Scotland supports higher temperatures in rooms that are unoccupied ' because it is considered that creating cold-spots is detrimental to the physical structure of the dwelling' (Scottish Government, 2012). In most cases, under-heating of rooms which are not used on a daily basis can lead to the development of damp, mildew and mould in the building fabric, on walls, wooden beams, floorboards, and most other surfaces. There is much more evidence supporting this likelihood now than there was at the time Scotland opted to include all under-occupied rooms in estimates of required fuel costs.

Given that damp, mildew and mould are also significantly associated with poorer respiratory health, it could be argued that under-heating practices can be equally detrimental to the health of a dwelling's inhabitants and not just the dwelling per se (Thomson, Sellstrom & Petticrew, 2013). Under-heating rooms can also mean householders close off parts of their home during the colder seasons, experiencing spatial shrink and reduced mobility. These adaptations can have adverse effects on both their physical health and their mental wellbeing.

Hence, as with the original fixing of indoor temperature regimes in Scotland, there was sound reason at the time to accept the need for all areas of the home to be heated in a relatively even manner. Since there is no new evidence to challenge this decision, but rather a significant accumulation of evidence in favour of it, it seems reasonable to suggest that rooms which may not be occupied very often in the colder seasons are nevertheless retained at temperatures recommended for bedrooms in frequent use. In this way, there is a dual opportunity to protect the house as well as the household.

There is however a tension between this perspective and current practice in advice to households about energy (and cost) saving which typically includes recommendations about 'zonal' heating. This means turning heating off or down in unoccupied rooms and only heating these rooms half an hour before they are going to be used. The UK's DEFACTO programme is trialling zonal systems, and reports that they are least beneficial for households who are at home most of the time. By contrast, they are most beneficial for families who are at work or school during the day (Beizaee et al., 2015). As illustrated on Table 4.7, there are substantial savings to be derived from using zonal controls in appropriate conditions.

Table 4.7.: Estimated gas use for heating by region.

Region (Weather station) Annual heating energy use CC a
( kWh)
Annual heating energy use ZC b
( kWh)
Reduction in heating energy use
NPV after 15 years: Basic system b
London (Gatwick) 15685 13839 11.8% £971
East of England (Hemsby) 15696 13848 11.8% £972
Northwest (Aughton) 15805 13936 11.8% £985
West Midlands (Birmingham) 16354 14379 12.0% £1,047
Northern Ireland (Belfast) 16374 14395 12.1% £1,050
Yorkshire (Finningley) 16507 14503 12.1% £1,065
Scotland (Aberdeen) 17346 15180 12.5% £1,160

a Calculated based on HDD base temperature of 17.8°C. For a typical weather year with heating months being October to April.

b Based on Department Of Energy and Climate Change ( DECC) energy & emissions projections central scenario for residential gas prices and discount rate of 5%.

This is an area where further discussion is needed across the policy domains of domestic energy efficiency and public health, and we return to the issue in Chapter 5.

Analysis of the Scottish House Condition Survey data in previous years (2008-2010) indicated ' a modest but significant correlation between under-occupancy and fuel poverty' (Scottish Government, 2012). However, when the analysis is updated (2013-2015) and further disaggregated, a rather different pattern emerges (see Table 4.8).

Table 4.8.: Under-occupancy by income poverty and fuel poverty 2013-2015 ( SHCS)

Underocc. Level Fuel + Income Poor Fuel Poor only Income Poor only Neither
N 000s N 000s N 000s N 000
% % % %
2 or more bedrooms underocc. 89 204 2 414
24% 46% 2% 27%
1 bedroom underocc. 136 140 20 598
37% 31% 28% 39%
Compliance with standard 129 97 40 477
35% 22% 56% 31%
Crowded 16 6 10 41
4% 1% 14% 3%

These data suggest that:

  • A quarter (24%) of those who are in fuel poverty and on low incomes reside in homes with 2 bedrooms or more under-occupied; this almost doubles (46%) for households who are fuel - but not income-poor;
  • 4% of those who are in fuel poverty and on low incomes are in homes which are overcrowded rather than under-occupied; this equates to 1% for households who are fuel - but not income-poor.

In other words under-occupancy among fuel and income poor households is relatively uncommon. The ' modest but significant correlation' between fuel poverty and under-occupancy is largely located in households on higher incomes.

Hence, under-occupancy does not seem to be a dominant feature of households in both income and fuel poverty. Given that there are health concerns related to under-heating any rooms that may only be used occasionally, there seems little justification for altering how under-occupancy is treated when estimating fuel poverty prevalence.

4.8. Definition and practice - tenuous links

Whether jurisdictions use a Boardman-based definition (Scotland, Wales and Northern Ireland) or the LIHC indicator (England), recent reviews of how alleviation programmes deliver in practice suggest that these are at best loosely guided by how the concept is officially defined. A review of health-related fuel poverty alleviation programmes suggest that neither definition was used to target resources: "it is as if different languages are being spoken at national and local level when it comes to how objectives are expressed and measured" (Fletcher et al., 2017). This perhaps reflects the extent to which how fuel poverty is defined in principle has become dislocated from how it is being addressed in practice.

In this context, the 2012 Evidence Review for Scotland noted that:

" The debates about fuel poverty have traditionally been maintained at quite a theoretical level whilst fuel poverty programmes on the ground attempt to operationalise fuel poverty into a practical concept using proxy variables to identify target households".

4.8.1. The trouble with proxies

In the past, determining eligibility for assistance from Scottish fuel poverty programmes has relied on proxies. These have most commonly been:

  • age of occupants;
  • location in an area of deprivation;
  • type/age of building;
  • rural location;
  • receipt of passport state benefits;
  • type of heating system;
  • modelled energy costs needed to attain a satisfactory heating regime.

Whilst these are all significant correlates of fuel poverty in Scotland, the strength of correlation seldom exceeds low-to-moderate (Mould, Baker & Emmanuel, 2014). For example, there is no statistically significant relationship between income as measured for the Scottish Index of Multiple Deprivation ( SIMD) and the distribution of fuel poverty at small area level based on Scottish House Condition data (Mould and Baker, 2017b). Furthermore, data for 2015 suggest that, of all households in receipt of those passport benefits which can trigger eligibility for state-funded fuel poverty assistance, only 20% are both fuel poor and income poor. By contrast, more than half of these households (54%) are neither fuel nor income poor.

Increasingly, proxies seem less than fit for the purpose of deciding who is eligible for state assistance via subsidised fuel poverty schemes:

" The groups prone to fuel poverty, as defined by [a needs to spend value] and income can only be reliably diagnosed by those vectors. Other risk factors [such as age, house condition, etc.] appear to be so diverse within the fuel poor population that they can equally apply to the population as a whole". (Scottish Government, 2012).

4.8.2. Replacing proxies with data Income

There are difficulties in obtaining accurate data on income, and these are substantial, but in many cases not insurmountable. At present many households being assessed for eligibility on fuel poverty programmes are referred for income maximisation checks. These are carried out by teams in the civil service who have access to HMRC and other income data. Could such checks be expanded to include all consenting households being assessed for eligibility and if so what resources would be required to achieve that? Initial feedback to delivery teams need only be in the form of a binary result on Income Poor/Not Income Poor and would disclose little of substantive sensitivity. For those who are Income Poor, actual income would then need to be set against required fuel costs, for which further consent could be sought. Modelled energy costs

Similarly, whilst most stakeholders would agree that required fuel costs are the metric of choice when estimating fuel poverty, since they capture needs to, rather than actual spend, a number of new databases on actual spend have recently become available in the public domain. These may be useful for identifying those most in need of state assistance.

The first derives from the National Energy Efficiency Data-Framework ( NEED), which has been established by BEIS to provide data on energy use and energy efficiency in Great Britain. It contains matched data on energy consumption and the energy efficiency measures installed in homes, as well as information on household characteristics and property attributes.

The second derives from the growing database available from smart meters in Great Britain. These meters provide detailed data on both electricity and gas consumption, available on a baseline of half-hourly intervals. Smart meter data can help identify customers likely to be in fuel poverty, particularly those who may be in deepest fuel poverty. For example, Figure 4.3. compares electricity consumption data from Northern Ireland's main retailer ( PowerNI) and illustrates a year's consumption for the population of all PowerNI customers as well as for a subset of their "high income" customers; the third group ("low income fuel poor" customers) are a subset in fuel poverty. Fuel poor customers show much less evidence of seasonal variation in electricity consumption, which suggests a finite expenditure capacity with little elasticity (Darby, Liddell & Hills, 2015).

Figure 4.3.: Monthly electricity consumption in Northern Ireland - consumption profile for all, high-income, and low-income/fuel poor customers

Figure 4.3.: Monthly electricity consumption in Northern Ireland - consumption profile for all, high-income, and low-income/fuel poor customers

Smart meter data are also valuable for locating vulnerable customers who have a regular pattern of self-disconnection or unusual diurnal patterns of under- or over-use:

" Smart meters could provide a useful additional means of identifying vulnerable customers and, in particular, patterns of heating use and bill payments suggestive of fuel poverty. Reducing the invisibility of the most vulnerable may need to be more prominent in models, as a potential public service benefit over and above energy savings, demand response, and ability to switch between tariffs and suppliers. Identifying and protecting vulnerable customers could, in the broader framework of consumer engagement, contribute significantly to acceptance of this technology." (Darby, Liddell & Hills, 2015).

In brief, the combination of three features in smart meter data:

  • low consumption;
  • low seasonal elasticity in consumption; and
  • frequent self-disconnection

can provide a valuable means for identifying households likely to be in greatest need of assistance. Energy retailers are already using smart meter data for this purpose, and anonymised data on consumption via smart meters in Scotland may have considerable value for identifying those most in need.

A final use of smart meter data is in the provision of more tailored and evidence-based energy efficiency advice to customers. Meta-analyses of smart meter data in this context indicate potential savings of 2% on a household's electricity bills through the direct feedback on consumption that in-house displays attached to smart meters can facilitate; with additional expert energy efficiency advice and support, meta-analysis indicates that this can more than double (to 5%) (Darby, Liddell & Hills, 2015). The CharloT project is an EPSRC-funded research programme in this area which uses specially designed sensors, alongside smart meters to collect data on internal temperature, external temperature, light levels, relative humidity and energy consumption. The data are sent via broadband and the mobile phone network to a server for capture and analysis, and are being used by energy agencies to develop household-specific energy advice packages (Fischer et al., 2017).

For these reasons, the potential of data being gathered by the National Energy Efficiency Data-Framework ( NEED) and smart meters in Scotland should be more fully explored, with a view to both finding more households who are in deepest fuel poverty, and identifying ways of assisting them and a much wider variety of households in reducing their energy costs.

Real data may also enable more extensive validity-testing of some key assumptions made in estimations of fuel poverty prevalence which derive from the UK's application of the BREDEM model. As noted by Herrero (2017) actual spending is well below modelled energy expenditure - most crucially in the higher income deciles, where there is less need for rationing or cutting costs; the ratio of actual to required energy spend tends to stabilise at 80% from income decile 7 upwards, which may suggest an over-estimation of household energy requirements in the BREDEM model at national level.

Conversely, many have argued that the BREDEM 2012 model (which is used in Scotland at present) underestimates how much energy costs in certain geographical areas, particularly in remote rural and island communities. Data are not collected on the actual price households need to pay for fuel, this being sourced from standard publications such as the Sutherland Tables and BEIS Quarterly Energy Prices. This is particularly problematic in the case of non-regulated fuels, such as heating oil, the cost of which can vary substantially from one geographical area to another. Whilst this problem is partially alleviated by assigning different price indices to the North and South of Scotland (prices for the former being higher), it is debateable as to whether this geographical differentiation is sufficient. As has been shown in Northern Ireland, the cost of purchasing 300 litres of oil varied across much smaller areas - from £175 to £216 based on concurrent survey data from retailers. This is a difference of almost 25% (see Figure 4.4.).

Figure 4.4.: Continuous price surface generated by spatial interpolation of heating oil prices in Northern Ireland. Data collected from suppliers (n = 119) in July 2011, January 2012 and July 2012.

Figure 4.4.: Continuous price surface generated by spatial interpolation of heating oil prices in Northern Ireland. Data collected from suppliers (n = 119) in July 2011, January 2012 and July 2012.

Whilst prices were among the cheapest around one of the ports into which oil is imported (Derry), it was among the most expensive around the other port (Belfast). Multivariate analyses suggested that prices were based on a complex set of interacting factors related to local market structures, supply costs, market competitiveness and socio-economic factors which affect demand (Walker et al., 2015). The same is likely to occur in Scotland, and could significantly underestimate the extent to which fuel poverty prevails in remoter off-grid areas. However, the extent of price variations, how readily these could be attenuated in Scotland, and how these variations might alter estimates of both prevalence and severity, remain unknown. The Panel thought that this was an area which could be relatively easily researched (and updated over time) by a trusted independent research agency.

4.8.3. Targeting those most in need: impacts on fuel poverty prevalence

If households in the extremes of fuel poverty are targeted for state assistance, many may not be taken out of fuel poverty altogether, even after the deepest possible retrofits. Such households would most likely still contribute to the prevalence of fuel poverty in Scotland. Of those that are removed from fuel poverty, the move could be ephemeral, in that a large proportion are likely to move from extreme to marginal fuel poverty [16] , which locates them in the so-called churn area; any increase in energy costs could readily move them back across the 10% required energy cost threshold and into fuel poverty again.

It has been argued that the headline "in fuel poverty" prevalence rates which regions of the UK publish could be construed as encouraging targeting on those just beyond the 10% threshold; for schemes to demonstrate maximal effectiveness, an optimal strategy would be to locate households with energy costs around 10.1% to 12% where modest retrofits will move them out of fuel poverty (Walker et al., 2014). Under such a scenario, there is potential for the proportion of households in extreme need to grow in prevalence.

If a strategy for targeting those most in need is to be pursued in the future, then that strategy may wish to focus on the prevalence of extreme fuel poverty, however fuel poverty is ultimately defined. Targets could then be set related to reducing that particular subset of fuel poor households. This will not only monitor outcomes on the basis of impacts on the target group, but will also ensure that positive impacts are not obscured by the very challenging target of removing all households assisted out of fuel poverty.

Targets are also likely to be more useful and informative if they are set (and outcomes published) annually. The original fuel poverty targets which all nations in the UK signed up to (eradicate fuel poverty where practicable by 2016) were set in 2001 - a 15 year target which afforded too much elasticity in the timeframe of failure, affording many years of negative publicity in which failure could be amply anticipated. It will be easier to avoid such a long tail of negative publicity with more short-term targets, and these will also afford more immediate opportunities to seek solutions and redress [17] .

4.9. Summary

The Boardman versus LIHC choice is not a simple matter of choosing between two definitions which differ from one another in nuanced and subtle ways. They are fundamentally different in how they measure the same construct. Weighing up their respective strengths and weaknesses, there is little to favour LIHC over Boardman or vice versa. Each definition has advantages and shortcomings when applied to data from Scotland, corroborating results from England and France.

Whilst adopting the LIHC indicator would more than halve fuel poverty prevalence, it should be noted that the exclusion of wealthier households from a Boardman definition would have a similar effect on prevalence, since the 2015 SHCS highlighted that more than half those in fuel poverty were not income poor. Using SHCS data averaged for the last three years (2013-2015), and Boardman's definition modified to exclude those who are not income poor, the proportion of Scottish households in fuel poverty would be 15%, compared with 34% using the Boardman definition as it currently stands (and 12% using the LIHC indicator).

But neither of these technical definitions can satisfy growing concerns about the place of fuel poverty in discourses concerned with energy justice and a fairer society in Scotland. At the time they coined their definitions, Boardman's specialism was largely in understanding heating in the housing stock, whilst Hills was an economist. Their definitions are technical, and in that sense they have been relatively immune to the vagaries of socioeconomic change or newly elected governments. This can be seen when comparing England's 2001 and 2015 Fuel Poverty Strategies. The first was published under Boardman during a Labour Government, the second under Hills and a Conservative/Liberal Democrat Coalition. As has been made clear in this report, the two definitions are radically different metrics with little cross-over. Yet the Strategies themselves show little evidence of footprint deriving from the different definitions in play. With the exception of the 2001 Strategy's focus on the goal to eradicate fuel poverty, the Strategies are remarkably similar: both emphasise partnership-working, both prioritise helping those most in need first, and in looking ahead, both identify the same impediments to achieving their goals.

Over time, both definitions have become increasingly remote from what teams delivering fuel poverty alleviation programmes have deemed most effective. The main concern with such a dislocation is that it becomes difficult to assess the extent to which funds invested in alleviating fuel poverty on the ground can be expected to alleviate fuel poverty prevalence as it is officially measured. For Boardman and Hills, the reasons behind such dislocation may have been different, since their definitions dominated in very different political and economic contexts. The UK Fuel Poverty Strategy of 2001 used a modified version of Boardman's definition to outline a Strategy that was positive, expansive and ambitious. It set out plans to "eradicate" fuel poverty by target dates, representing this issue as one which could be systematically whittled away.

The 2011/12 review took place against a much gloomier backdrop:

  • failing targets for eradicating fuel poverty;
  • a 2008 judicial review in which Friends of the Earth and Help the Aged challenged the view that Government had done everything reasonably practicable to eliminate fuel poverty in England; the challenge was unsuccessful, and significantly dented the morale of local authority teams and other stakeholders;
  • a shrinking Government budget for tackling fuel poverty in England: between 2009 and 2012, this was cut by almost one-third (31%) (Jansz & Guertler, 2012);
  • the progressive dismantling of England's flagship fuel poverty alleviation programme (Warm Front), which was completed in 2013.

The Hills LIHC indicator had relatively little potential to make a substantial impact on government fuel poverty programmes in England, in part because these were being so systematically scaled down and deprioritised.

In this context, it is perhaps worth noting that the current review of a definition for Scotland is taking place against the backdrop of a fuel poverty budget which has increased significantly between 2008/9 £66.9M to £114M [18] , and in which energy efficiency was recently classed as part of a national infrastructure priority. [19] There is now, perhaps, greater potential for a renewed synergy between definition, Strategy, policy and programmes than had hitherto been possible.

If a new definition is sought in Scotland, and the intention is to encompass many of the broader aspects of the fuel poverty agenda, then the search for a new definition will have to encompass changes to:

  • whose voices are heard;
  • how fuel poverty is measured;
  • who is prioritised for assistance;
  • what steps should be taken to alleviate it;
  • what outcomes are monitored.

In other words, the search for a new definition could require root and branch reform.

4.10 Last word - lay definition

Finally, sight should not be lost of the value of having a simple lay definition of fuel poverty running alongside whichever technical option is chosen. In line with the increasing discourse on fuel poverty and inequalities, most of which derives from UK scientists (e.g. Gillard et al., 2017), it seems increasingly likely that this will encompass concepts such: as affordability, justice, equity, community empowerment, public health and human wellbeing. It need not be laconic, since it should do justice to the complexity of the causes and consequences of fuel poverty, but it should ideally be written in a manner which makes this complex construct more accessible and easy for people to understand.

Key Conclusions on the Boardman (1991) definition and the LIHC indicator

There are strengths to be found in both the current Scottish definition (based on Boardman) and the current English definition (based on Hills).

However, Boardman's definition does not confine fuel poverty to households on lower incomes and as such it does not align with Section 95 of the Housing Act which indicated that 'a person lives in fuel poverty if that person is a member of a household with a low income living in a home which cannot be kept warm at a reasonable cost'. (bold font added)

In a similar fashion, the Hills LIHC indicator excludes many households from being considered fuel poor, despite the fact that they may be on very low incomes indeed. There are likely to be practical forms of assistance which could significantly reduce the burdens arising from their energy bills. Where the meaning and significance of being fuel poor are a consideration, the burdens associated with their energy costs support the view that this group should remain an integral part of those deemed to be in fuel poverty.

Furthermore, the headline prevalence estimates of fuel poverty in Scotland under an LIHC indicator are insensitive to changes in fuel prices, which means that second-tier data (on the fuel poverty gap) needs to be accessed before any understanding can be gained of what such changes imply for prevalence.

These core drawbacks alone point to the need for a different definition in Scotland.

However an additional and salient drawback with both of the conventional options lay in the fact that neither of these definitions currently bears a substantive relationship to how fuel poverty programmes are delivered on the ground.

The Panel accepted emerging consensus around the idea that "an affordable heating regime" is only one aspect of a healthy indoor climate; optimal indoor conditions should, we thought, also encompass aspects of ventilation, condensation, mould growth, and damp.

The Panel had concerns about the use of inaccurate proxies for estimating fuel poverty - whether for national prevalence data or on the doorstep. Wherever possible in the medium-term, efforts to replace these with more accurate data (particularly regarding income and energy costs) were strongly supported.


Back to top