Characterizing the energy use of disabled people in the European Union towards inclusion in the energy transition

Designing environmental policy to take account of social difference is increasingly recognized as essential to address both effectiveness and justice concerns. So far there is limited research on the experiences of disabled people in the environmental literature, amounting to a failure to recognize this substantial constituency. Here we compare disabled households’ embodied energy use, income, risk of poverty and energy poverty, and other socio-demographics with other households in the European Union. We find that households including an economically inactive disabled person earn less and consume 10% less energy than other households, and are more likely to experience energy poverty. Disabled households have lower consumption than other households in most categories, with the exception of basic consumption such as food, energy at home (gas and electricity), water and waste services: in effect they have less—and sometimes inadequate—access to resources. We conclude that more attention should be paid to disabled households needs to ensure a just energy transition. A just energy transition requires consideration of the needs of all people, yet disabled people are often overlooked. Ivanova and Middlemiss analyse the energy use of disabled households in the European Union, finding that on average they consume less energy and are more likely to experience energy poverty.

R ecent work on sustainable consumption argues for a more nuanced approach to policy, taking into account a range of social differences when planning and delivering interventions [1][2][3][4][5][6] . Given the substantial differences in consumption levels associated with affluence 7 , the wide variation of energy use in wealthy nations more generally 8,9 and the differential impact of the effects of climate change on different people, it is important to understand the differences in what people are able to consume when considering how to address a transition to a more sustainable future. In the emerging literature on social difference, scholars pay attention to populations that experience injustice, pointing out that the interests and experiences of women, people from ethnic minorities, working class and disabled people are poorly accounted for in environmental policy and research. Taking account of social difference in environmental policy and practice is both a matter of effectiveness (to avoid addressing people with policies that they will not be able to engage in) and justice (offering fair options for people who have particular needs).
So far there is very limited research on the needs and experiences of disabled people in the environmental literature 4,[10][11][12][13][14][15] . Note that we have chosen to use identity-first language here, both in naming these disabled households and referring to disabled people. This is the way in which disabled activists refer to disabled people in the United Kingdom, where we are based; we wish to support their choice in this and do not intend to cause harm (see https://disabilityunion.co.uk/ person-first-or-identity-first-the-importance-of-language for more information). Disabled people are subject to a failure in recognition justice: they are largely invisible in environmental policy and practice, and rarely discussed as having particular needs or facing particular challenges in a transition towards a more sustainable future. This is despite the fact that disabled people are regularly supported by their governments in developed nations (including European Union (EU) nations) and seen as important targets of social policy. When we consider that of the 446 million people living in the EU, around 100 million are believed to be disabled 16 , this is by no means a rare experience, and disability is a topic that merits more attention from environmental scholars and policymakers. Disabled people may also be more vulnerable to climate change consequences such as extreme temperatures or emergency relief being inaccessible 12 .
Disabled people's experiences are diverse and disabled people will live very different lifestyles [17][18][19] . The concept of disability is a very broad one, and can include a wide range of impairments such as autism spectrum conditions, long-term health conditions, mental health conditions, physical or mobility impairments, sensory impairments (for example, deafness and blindness) and learning difficulties 20 . Although some disabilities are life-long, others develop during the course of life and, as such, can present challenges for everyday living as people have to adapt to a new impairment and the changes in lifestyle that this necessitates. When considering transitions to less environmentally impactful futures, an inclusive approach will involve understanding the diverse impacts that disability has on people's energy use and requirements.
There are, however, some common trends in disabled people's lives that are well recognized and will have clear energy implications. Disabled people on average earn less than non-disabled people 18 , and their energy use is likely to be lower than average given that income is by far the primary driver of carbon and energy footprints 7 . Some disabled people will find it harder to access employment, education and social opportunities 21 , which will also impact on their energy use. People in poor health certainly travel less 22 and, as such, are likely to have higher requirements for energy in the home. Some disabled people will have specific needs to use energy: higher room temperature, use of electrical equipment or access to a car 21,23 . People with disabilities tend to have higher health-care expenditure and health needs 18 , yet they are also less likely to access health services and are more likely to have worse health and wellbeing outcomes 18 . Studies on disability and energy poverty show that a combination of low income levels and high energy needs results in more disabled people being unable to afford adequate energy services, often exacerbating their disability as a result 21,23 . Furthermore, there is a strong connection between people in poor health and energy poverty in most EU nations 24 .

Characterizing the energy use of disabled people in the European Union towards inclusion in the energy transition Diana Ivanova ✉ and Lucie Middlemiss ✉
Designing environmental policy to take account of social difference is increasingly recognized as essential to address both effectiveness and justice concerns. So far there is limited research on the experiences of disabled people in the environmental literature, amounting to a failure to recognize this substantial constituency. Here we compare disabled households' embodied energy use, income, risk of poverty and energy poverty, and other socio-demographics with other households in the European Union. We find that households including an economically inactive disabled person earn less and consume 10% less energy than other households, and are more likely to experience energy poverty. Disabled households have lower consumption than other households in most categories, with the exception of basic consumption such as food, energy at home (gas and electricity), water and waste services: in effect they have less-and sometimes inadequate-access to resources. We conclude that more attention should be paid to disabled households needs to ensure a just energy transition.
Given these trends, we can expect that disabled people will use less energy, but this is not necessarily through choice: indeed it is likely to be at the cost of disabled people not having their needs for energy and other resources met. Energy is not purchased for its own sake, but for the energy services that it delivers 25 ; some energy services are essential for wellbeing and survival as in the case of cooking, heating and access to health and education infrastructure 8 . Although access to adequate energy services has clear wellbeing implications, research exploring the energy use of disabled people and the implications for energy poverty is scarce. In this study we provide a comprehensive quantitative analysis of the energy use of disabled households in the EU in various consumption domains including transport, housing, food, health services and other consumption. We use the term disabled household throughout to mean a household with at least one disabled and economically inactive member. This category captures people unable to work due to a physical or other disability and excludes disabled people who are economically active or who fit into another category of economic inactivity such as children, retired or in full-time education. Given that economically inactive disabled people are more likely to have more substantial impairments, and to suffer from material deprivation and poverty compared with employed disabled people [26][27][28] , our analysis of energy use, poverty and other social indicators among economically inactive disabled people in the EU is still extremely policy relevant, even if not representative of all disabled people. We consider both home energy (where use of energy in final consumption can help perform actions directly such as lighting, cooking, heating and cooling) and energy embodied in consumption (which is required to provide goods and services such as food, clothing and communication) 25 .
In this study we quantify household energy use on the basis of household consumption data from Eurostat's household budget surveys (HBSs) and energy intensities from EXIOBASE. Both data sources reflect trends in 2010, which is the latest available Eurostat wave at the time of submitting this article. First, we explore the energy use of disabled households in comparison to other households, highlighting differences in consumption across various consumption categories. Through a multivariate regression analysis, we estimate that disabled households have 10% lower energy use, which is associated with lower consumption of transport, health and  recreation services, hotels and restaurants, machinery and equipment, housing and others. Second, the differences in energy profiles are coupled with lower income levels and higher rates of poverty and energy poverty. These signal a striking inequality, exclusion and inadequacy of living conditions that is rarely attended to in environmental policy. We conclude by highlighting the need to take into account disability issues when planning the energy transition and provide recommendations to enhance environmental policies in this regard.

Energy consumption trends of disabled and other households.
Disabled households have consistently lower income and consumption levels than other economically inactive and economically active households in the EU in 2010 (Fig. 1a,b). On average, disabled households have 76% of the income of other economically inactive households and 60% of the income of economically active households; 33% of disabled households are in the lowest income quintile, compared with 21% and 17% for other economically inactive and active households, respectively. Furthermore, only 6% of the disabled households are in the highest income quintile, compared with 28% for economically active households ( Supplementary  Fig. 7). The lower incomes across disabled households have clear implications for decent work, purchasing power and social equity 29,30 , where the inequality of disability is further exacerbated by income inequality 29 . The lower incomes and higher reliance on disability, retirement and other government benefits among disabled households ( Supplementary Fig. 8) further highlight the lower financial resilience and ability to invest in alleviating energy poverty (for example, through energy efficiency or micro generation) among disabled people. The lower incomes further translate into lower energy use among disabled households (a two-sided Spearman's correlation coefficient of 0.58, P = 0). The differences in energy use between household types are smaller than income, though still substantial (Supplementary Table 4). This probably reflects higher energy requirements for disabled households. On average, disabled households in the EU note about 91% of the energy use of other economically inactive households and 77% of that of economically active households, at 0.10, 0.11 and 0.13 TJ per capita, respectively (Fig. 1c). This is equivalent to 28, 31 and 36 MWh per capita, respectively. The substantial energy inequality across countries ( Supplementary Fig. 4) and income levels ( Supplementary Fig. 2) is alarming, as energy is central to nearly every major social challenge and opportunity of our time (SDG7). The distribution of energy use is interlinked with achieving decent living 31,32 , alleviating poverty 33,34 and providing for basic needs around nourishment, health, education and shelter 34 . We find Belgium, Czech Republic, Slovakia, Luxembourg and Denmark to have the lowest relative energy use gap, where the energy use of disabled households still amounts to 78-88% of the energy used by economically active households. The highest difference occurs for Romania, where disabled households have energy use amounting to 47% of that of economically active households ( Supplementary Fig. 4).
Disabled households have higher shares of energy associated with food, tobacco, gas and home fuels, electricity, waste and hot water services, and lower shares of energy use from health and recreation, motor fuels, public and air transport, rent, machinery and equipment, hotels and restaurants, and other consumption (Fig. 1d). This is consistent with our expectations that household consumption of disabled households will be shaped by differing access and needs.
The share of households living at risk of poverty and energy poverty (defined as households below the national median income, spending more than 10% of household income on energy costs) is also higher among disabled households at 24% and 17%, compared with 16% and 12% among other economically inactive households, and 11% and 6% among economically active households (Fig. 2). These levels are consistent-though slightly lower-compared with EU statistics estimating the risk of poverty or social exclusion at 30% among people with some or severe activity limitations in 2010 35 . The risk of energy poverty is lowest among disabled households in countries such as Cyprus, France and Luxembourg at 6-8%; even in these countries, however, the average risk of energy poverty is higher among disabled households than other households ( Supplementary  Fig. 6). These findings contribute to prior evidence suggesting that energy poverty can be several times higher than the average for certain social cohorts, for example in single parent households mostly headed by women 21,36 .
The average per capita energy use is similar among the disabled and energy poor, disabled and not energy poor, and other energy-poor households where both disability and energy poverty are linked to lower energy use (Fig. 3a). A comparison within disabled households suggests that disabled and not energy poor households have higher energy use associated with food, health, insurance and recreation, motor fuels, public transport, machinery and equipment, and other goods and services (Fig. 3b). Although the distribution of energy changes, the total energy amount used is very similar on average. This suggests that reducing energy poverty among disabled households (for example, through investment in energy efficiency) will not necessarily lead to increases in the overall energy use but would rather allow for more adequate satisfaction of energy needs. Among energy-poor households, home energy use (gas, electricity, solid and liquid fuels) still comprises a higher proportion of the total energy use of disabled households compared with other households, at 40% and 36%, respectively. This suggests that the effects of disability and energy poverty together exacerbate further the potential to achieve energy requirements for decent living in non-housing consumption domains.
Other socio-demographic trends. Income, household size, population density and age have been shown to have important implications for energy use and associated GHG emissions 7,37,38 . Economically inactive disabled people are less likely to live on their own with a share of one-person households of 25% compared with 36% for economically active households (Fig. 4a). This trend is particularly strong in countries of Central, Eastern and Southern Europe ( Supplementary Fig. 9). At the same time, disabled households have higher shares of large households (five or more members) at 8% compared with 3.5% for economically active households. The bivariate regressions statistically confirm the positive association between disability and household sizes across most EU countries, which could be due to family support. That way, disabled households report higher potential for household sharing and so reduced energy use at home 38 . Furthermore, disabled households in the EU are less likely to be located in densely populated regions with a share of 41% compared with 51% for economically active households (Fig. 4b). This difference has clear implications for sustainable transport policy, which should be designed in a way that does not overburden disabled households, residents of which may already suffer from reduced accessibility and mobility. Disabled households have a greater representation of people between 45 and 59 years (36%) and people above 59 years (20%) than other households (Fig. 4c). Figure 5 depicts the coefficient plot on the effects of disability in the European sample. The dependent variables change in the models: we depict the effect of disability on the total or category-specific energy use per capita in logarithmic form. Holding income, household size and other socio-demographics fixed, we find that disabled households are associated with 10% less total energy use than other households. The regression model has an adjusted R 2 of 42%, based on more than 152,000 observations (Supplementary Table 5); thus, in addition to their lower incomes, disabled people experience additional effects around inclusion and opportunity to live fulfilling lives, which come with energy implications.

Multivariate regressions.
In addition to differences in socio-demographics such as lower income and higher household size, disabled households are associated with reduced energy use in various consumption domains. Disabled households have 15% lower energy use associated with health, insurance and recreation services. Controlling for income and other socio-demographics, disabled households use less home energy: 2% less energy use associated with gas, liquid and solid fuels, and 16% less energy associated with housing rent and real estate services. Disabled households in the EU also use much less transport energy at fixed socio-demographics, namely, 14% less energy from motor fuels and 16% less energy from public and air transport. Furthermore, disabled households report less energy use associated with machinery and equipment (10%), hotels and restaurants (21%), and other goods and services (10%).   provides additional cross-country and socio-demographic detail for the regression model with the total per-capita energy use in logarithmic form as a dependent variable. The table shows a wide variation in the disability effect across EU countries. In countries such as Bulgaria, Czech Republic, France, Greece, Croatia, Hungary, Ireland, Latvia and Poland, disabled households have lower total energy use, even at constant levels of income, household sizes and other socio-demographics. The lowest energy use among disabled households in these countries are noted for health, insurance and recreation services (up to 40% less energy), gas, liquid and solid fuels (up to 14%), electricity (up to 20%), motor fuel (up to 30%), housing rent (up to 26%), machinery and equipment (up to 27%), hotels and restaurants (up to 26%), and other consumption (up to 27%). The energy profiles of disabled households further point to a lower engagement in socially visible leisure consumption than other households, with systematically lower consumption of recreation, hotels and restaurants, and travel services. It is important to note that in this section we discuss energy use differences between disabled and other households at fixed income, household size and other socio-demographics; however, as shown in Figs. 1a and 4, there are substantial differences in the socio-demographics between disabled and other households, which have additional implications for energy use. For example, Table 1 depicts a strong positive association between income and energy use across all EU countries (for past evidence on this effect, see refs. 7,8,38 ). In the total sample, a doubling of income produces an energy use increase of 46% (Table 1). The lower income levels among disabled households thus further contribute to their lower energy use compared with other households. Ensuring more equal opportunities of education, employment and income for disabled households directly may help equalize energy use-and so the access to essential energy services. Furthermore, the energy consumption associated with health, insurance and recreation, rent and other consumption increases most with rising income across the consumption categories in the disabled sample (Supplementary   Multivariate regressions with a dependent variable: the log of the total energy footprint per capita by consumption category. Independent variables: disability in the household, log of net household income, household size, urbanity, age of household members and geographical region controls. As the dependent variable is in logarithmic form, the exponent of the coefficients is to be interpreted. Sample sizes: total, 152,425. Supplementary Table 5 provides the coefficient estimates, standard errors, t-and P values, total sample sizes and adjusted R 2 across consumption categories. Table 8). The income coefficient is insignificant in the context of public and air transport energy use for the disabled sample, even though it is highly significant for the total sample (Supplementary Table 5).

Discussion
Disabled households are more likely to be experiencing energy poverty, and have lower energy use and higher energy needs than other households ( Table 2). The disparities in energy use are substantial, showing a deep-and as yet rarely recognized-inequality in access to essential energy services. Past evidence also suggests that failing to satisfy energy needs may cause pain or further exacerbate the disability 3,23 and the observed inequality.
These insights are particularly important in the context of the substantial EU policy efforts known as the energy transition that address the transition to a low-carbon future. Energy transition policy shows an increasing recognition that vulnerable energy consumers (defined as the energy poor) must be identified and appropriately addressed by national government policy 39,40 . There is a recognition here that not all energy consumers have the same needs, and that different consumers have different resources available to address these needs. At present, disabled households are largely invisible in this policy effort or at the most addressed only as part of the larger energy poor category. Although energy transition policy clearly needs to pay closer attention to energy poor households, our work suggest that disabled households might need additional support and consideration even if they are not energy poor.
At the least, energy transition policy needs to take account of the needs of disabled people, who may be at risk of being further marginalized by policies that require households to reduce consumption or increase investment. Disabled people may face multiple physical, financial, organizational and social barriers to participation in pro-environmental activities 13 , which tend to go unrecognized in policy appeals towards active, choice-making individuals 4 .
We hope that this paper offers a useful step in recognizing disabled people as an important demographic for EU energy policy, and in documenting the disparities in energy use between disabled households and others. Given these disparities, there is also a need to assess the impacts of energy and climate policies on this substantial demographic, and in doing so to understand the distributive consequences of policy. This requires political commitment to monitoring impacts of policy on disabled people, as well as promoting the inclusion of disabled people in policy-making. Increasing the representation of disabled people 13 in the design and implementation of environmental and energy policies will probably help avoid regressive impacts and better capture their potentially higher energy (and thus emission) requirements. A wider recognition of disabled people in environmental citizenship debates is also key to highlight their agency and enable a contribution and fulfilment of responsibility that disabled people can and wish to make 13 .
The specific challenges that disabled households are likely to face according to our findings above, are associated with these households's lower incomes, and higher consumption of particular energy services. These include a sensitivity to increases in living costs, particularly to costs of home heat and electricity, given that they consume proportionately more of these. Disabled households are also less likely to have capital to invest in the changes associated with the energy transition. Note that these characteristics paint a picture of the average experience of a disabled household in the EU, there will of course be considerable variation between nations and between households. This represents an opportunity for research: to investigate the range of experiences of disabled households in the EU and to identify national or regional contexts in which the conditions are conducive to a more positive energy consumption experience. For example, countries such as Denmark, Slovakia, Belgium, Czech Republic and Luxembourg show lower energy inequality across household types in our analysis, and in general report lower risks regarding poverty, in-work poverty and social protection 26 . Total sample -0.10*** 0.46*** -0.11*** -0.11*** -0.08*** 0.00 0.03*** 0.04*** 0.05** Dependent variable: the log of the total energy footprint per capita. Independent variables: disability in the household, log of net household income, household size, urbanity, age of household members and geographical region controls. Italy excluded due to missing income data. Two-tailed tests. We present the standard errors and P values in Supplementary Table 7. Significance levels: *P < 0.05; ** P < 0.01; *** P < 0.001.
With many nations aspiring towards a transition from gas to electric heating, and to dramatically increase the energy efficiency of the housing stock, the consideration of disabled households's needs in these policies will be critical. Different nations will put different amounts of the cost burden of this transition on the consumer according to what is politically acceptable, which is likely to have more substantial negative effects for disabled households 41 . Changes to travel costs and infrastructure will also affect disabled people, despite their relatively low consumption of these services at present. Given the constraints on energy use in disabled households that are visible in our data, the energy transition clearly risks further restricting access to energy-and carbon-intensive necessities for disabled people.
There may also be opportunities to address disability and energy policy in conjunction. For instance, given that disabled households have relatively consistent home energy (gas and electric) use across energy poor and non-energy poor households, this suggests that disabled households would be a good candidate for investment in household energy efficiency. This would need substantial financial support (no cost to household measures) given that most disabled people are on low incomes. It would also likely have more substantial health benefits (on average) than energy efficiency measures for non-disabled households. Improvements in social provisioning are also key to reduce income inequality and enhance social support and, in turn, life satisfaction 30 . Calls for universal adult social care for all in need emphasize that high-quality care is essential to maintain health, autonomy and capacity for social participation, which can simultaneously improve social wellbeing, enrich local communities by enabling a wider participation and prevent or delay adverse health effects 42 .
Our findings, and the invisibility of disabled households in the energy transition, suggest the need for a closer relationship between disability and energy policy. Disabled people are very disadvantaged in terms of energy and other social indicators in our data. Clearly, there is room for more radical approaches to level up the gap between disabled and non-disabled households. Italy provided data on consumption expenditure and disability as a current activity status, but no income data. The National Statistical Institutes in each member state carry out the HBSs with the main focus on consumption expenditure, where each country has considerable freedom to decide on the methodology and resource allocation for the respective survey 43 . Eurostat then harmonized and disseminated HBSs for a total of 174,966 households among these 19 countries. The surveys reflect trends in 2010, which is the latest available Eurostat wave at the time of writing this article.

EU
The HBSs do not collect information on disability of household members directly. They instead collect data on the current activity status of household members, including a permanently disabled option for economically inactive individuals 44 , where individuals can be classified as permenantly disabled or unable to work due to a physical or other disability. This definition excludes disabled people who are economically active (for example, working or unemployed) or economically inactive while fitting any other activity status (for example, children, the retired or those in full-time education). Although we are certainly very interested in understanding their energy use patterns, in the current analysis we do not have the data to single out the energy patterns of disabled people that are economically active or economically inactive under another activity status. The subset of persons who are economically inactive due to disability on which we focus on here is thus not representative of the population of disabled persons (see Supplementary Table 1 for other statistics on disability). Supplementary Table 2 contains the sample distribution by household type.
Other surveys show that the EU employment rate among disabled people amounts to 47-51% (compared with 67-75% among the non-disabled) 26,45 , that is, about half of disabled people are unemployed or inactive. Although we cannot generalize our findings to all disabled people or capture in-work poverty, the Public and air transport (-) Past evidence suggesting lower car, air, active and public transport associated with poor health status 22 .
Home energy No significant differences in the regression analysis, but higher gas and electricity shares than other households (+) Heating and cooling requirements (+), for example, higher indoor temperatures, longer periods of warmth and air conditioners use 3,23 Past evidence suggesting higher electricity use associated with poor health status 22 .
Home appliance use (+), for example, additional washing and drying facilities 3 , higher costs associated with pre-payment meters 23 Food No significant differences in the regression analysis, but higher food shares than other households (+) Greater risk of malnourishment 18 Health and recreation  Table 6).
subset of economically inactive disabled households is likely to be among the most disadvantaged and excluded. Economically inactive disabled people of working age are more likely to need assistance and face higher risks associated with more severe levels of impairment and barriers to social integration than employed disabled people 26,46,47 . Economically inactive disabled people are also more likely to suffer from poverty and material deprivation (including inability to pay energy and other household bills), particularly in countries with lower social transfers 27,28 . Furthermore, austerity measures enacted by governments after the financial crisis of 2008 and more recently the COVID-19 pandemic disproportionately impacted the lives of disabled people and targeted cuts in disability-related benefits for working age people 28,48 . Consequently, there is a rationale for focusing our analysis on a subset of economically inactive disabled people who are of working age and outside the labour market primarily due to disability. In our analysis, we compare disabled households with (1) other economically inactive households, which includes households where at least one person of legal working age is economically inactive due to retirement, training or unpaid work experience, fulfilment of domestic tasks, military or community service 44 ; and (2) economically active households, where none of the household members of legal working age is economically inactive, including working individuals and the unemployed seeking work 44 .
Energy use calculations. We used the multiregional input-output database EXIOBASE v.3.7 for 2010 energy footprint analysis. We applied a net energy indicator measuring annual energy use in terajoules at the product level (200 products). Net energy is suitable for footprint analysis as there is no double counting associated with the conversion from primary sources (derived directly from nature, for example, coal) into secondary sources (transformed for a certain industry or household use, for example, electricity) 49 . For the purpose of this article, we use energy footprint and energy use interchangeably.
Net energy accounts depict the energy use of households and industry (to produce products for household consumption) as well as losses associated with conversion and distribution processes. In terms of final consumption categories, our net energy accounts include both home energy and energy embodied in the consumption of goods and services delivered through the global supply chains. That way, our analysis goes beyond traditional accounts of home energy (for example, electricity use and heating requirements) and encompasses further energy uses, which are key to the provision of wellbeing (for example, food, mobility, health services, equipment, education). For the purpose of our analysis, we communicate the total net energy, which combines transport energy, non-transport (industry use and other) energy, non-energy use and losses, and energy industry's own use. Non-energy use covers fuel use as raw materials in the different sectors and are not consumed as fuel or transformed into another fuel 50 such as petroleum products and coal. Energy data were aligned with the residential principle and consolidated with the International Energy Agency energy flow and classifications to enable allocation to EXIOBASE industries, final consumption and products 51,52 . Modelling uncertainties arise from a lack of energy and extraction detail, allocation of household direct energy use, hidden or confidential data and the conflict between the residence versus territorial principle 53 .
We harmonized the consumption data from the HBS and energy extensions from EXIOBASE to calculate energy use for our household level sample. Supplementary  Fig. 1 details the calculation procedure and the harmonization of sources.
Additional social indicators. Energy-poor households are unable to afford the level of energy services needed for a decent life 54 . Here we categorize households as living in energy poverty when they spend more than 10% of their household income (or total consumption spending for Italy and Luxembourg) on energy costs at an income level below the national median level 9 . This amounts to an adapted 10% measure, counting households as energy poor when they spend 10% or more on energy costs and they have an income level below the national median level. In measuring energy poverty this way, we combine the UK's old and new energy poverty definitions 54 . To be specific: there are two differences between our indicator here, and the original 10% measure proposed by Brenda Boardman. (1) We only include households that earn below the national median level (thereby excluding wealthier households who are overconsuming); (2) we use real rather than modelled energy costs (this means that we are likely to lose some of the underconsumers who are self-disconnecting: so called hidden energy poor). This is the best available indicator in the data set we are working with here, although we recognize that it does not necessarily reflect the full picture of Energy Poverty (among disabled people) in each nation.
The population living in energy poverty and poverty is largely overlapping as they share common drivers such as low incomes, high fuel bills and ill health 54 .
Here we measure risk of poverty and social exclusion through the Eurostat's threshold set at 60% of the national median disposable income 55 .
The statistical model. We present income, consumption, energy use and poverty bar charts, graphing the mean and the 95% confidence interval for each household type using error bars 56 . Similar to a past energy analysis 38 , we perform a multivariate regression analysis for the total sample and for each country c separately; yet, this time we include disability as an independent variable in the regression model, which has not been done to our knowledge. The unit of analysis is the household i, where we estimate household energy use based on household consumption data by-product and country-level carbon intensities by-product, where beta is the regression coefficient varying by variable and country.
LNE stands for the estimated energy use measured in terajoules per capita, in logarithmic form, total or across consumption categories. DISABLED captures households with disabled and economically inactive persons (dummy variable). LNINCOME measures the effect of net disposable income per capita, in logarithmic form, which can be interpreted as the income elasticity of total energy use. HHSIZE measures the number of household members, sharing accommodation and expenses. The population density dummies DENSE and INTERMEDIATE divide the sample on households living in cities with at least 500 inhabitants per km 2 , households living in towns and suburbs with a density between 100 and 499 inhabitants per km 2 , and sparsely populated areas with less than 100 inhabitants per km 2 (see past documentation on the classification for more details 38,57 ). The AGE variables measure the role of age distribution for consumption and energy use, particularly depicting the number of household members within the following age categories: below 14 years old; between 15 and 29 years old; between 30 and 44 years old; and between 45 and 59 years old (with above 59 years olds as a base variable). The model further includes a set of geographical REGION variables accounting for regional differences such as technological, geographical and socio-economic context, which may influence energy use (categorical variable). The regional variable uses NUTS 2010 two-digit level unless otherwise specified (Supplementary Table 3). We add another model specification including other economic inactivity as a variable, which does not significantly affect the effect of disabled and economically inactive ( Supplementary Fig. 12); in essence, we still find a significant difference between disabled and other economically inactive households holding income, household size, density, region and age distribution constant.

Limitations.
Although the EU's HBSs are largely harmonized 43 , substantial differences occur in terms of detail that surveys collect and disseminate. More harmonized and nuanced definitions of disability and further details on what kind of impairment or disability people will allow for a more representative assessments of energy consumption and needs among disabled people and more nuanced interventions depending on the type of disability.
In the modelling of energy use, our analysis captures household consumption and the energy embodied in household purchases. We therefore cannot capture differences across household types and countries in terms of consumption of community and government services (and their associated impacts) that do not result in household expenditure outflow. Furthermore, more detail on the physical, socio-economic and political context in which energy use takes place will be beneficial to elaborate on the circumstances in which energy demands associated with disability increase or decrease. We also could not account for price differences between disabled and non-disabled households in our analysis; for example, in England, a greater proportion of disabled households are on pre-payment meters compared with other households, associated with challenges such as increased costs, the risk of self-disconnection and the need to top up 21 . If disabled households systematically pay higher prices, we will be overestimating their energy use compared with non-disabled households.
Although our model lacks a temporal dimension, we expect that the observed trends between disability and energy use are dynamic in several ways. On a micro level, the dynamics around the course of the disability may be key, where associated energy use is not necessarily static and may change, for example, with high quality care 18 . On a macro level, factors around energy prices changing with the energy transition, increased climate impacts and frequency of extreme events and changes in the broader political, infrastructural and provisioning context may all have key energy implications and so influence the living situation for disabled people.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The energy use data behind the analysis is openly available through the Research Data Leeds Repository hosted by the University of Leeds at https://archive. researchdata.leeds.ac.uk/891/ (ref. 58 ). The dataset contains energy footprints (total and by consumption category) and categorical variables on poverty, energy poverty and disability, all of which are generated by the authors. It also contains household IDs, country codes and household weights generated by Eurostat and corresponding to the HBSs. The household and country ID variables are included to allow for households to be identifiable by users of the HBS dataset disseminated by Eurostat. Other variables in the HBSs (for example, household size, income, population density and so on), which have been used in the statistical analysis, are subject to controlled access through the official Eurostat portal. Access to the HBS microdata is granted for scientific purposes only through applying to have your research organization recognized as a research entity and submitting a research proposal requesting the specific data. Once the research entity is recognized (mailbox for requests to be recognized as a research entity: ESTAT-ENTITIES-ASSESSMENT@ec.europa.eu), the research proposal is submitted via an online form via Microsoft Access Workflow Tool (accessible via an EU-login). The national statistical authorities are consulted before approving research proposals, where the standard consultation period is four weeks. Access is usually granted to partially anonymised data for a period of time following the guidelines for publication. More information about the procedure can be found at https://ec.europa.eu/eurostat/documents/203647/771732/How_to_apply_for_ microdata_access.pdf/82d98876-75e5-49f3-950a-d56cec15b896. The EXIOBASE database is available at https://zenodo.org/record/4277368.

Code availability
The codes (to estimate energy use based on household expenditure (MATLAB), and to perform the statistical analysis (Stata) and concordances used to derive final energy use, are openly available through the Research Data Leeds Repository hosted by the University of Leeds at https://archive.researchdata.leeds.ac.uk/891/. The energy use data behind the analysis is openly available through the Research Data Leeds Repository hosted by the University of Leeds at https://archive. researchdata.leeds.ac.uk/891/ (ref. 58 ).