Modelling Decent Living Energy
The approach used to estimate global energy requirements is bottom-up, and involves combining activity levels for each material dimension of decent living with associated energy intensities, before summing across dimensions to obtain total final energy consumption. For example, for residential buildings, we have direct energy intensities for heating and cooling and indirect intensities of construction, all in MJ/m2, which can be multiplied by the assumed activity-levels, in m2/capita, to obtain energy use. The estimates thus include both direct energy use and the indirect energy required to produce products and infrastructures; the latter is divided by product/infrastructure lifetimes to give annualised values of indirect energy consumption. The calculations thus involve simple multiplication and summation. However, compiling the input data is an in-depth process.
The minimum activity-levels assumed in the scenarios are intended to describe what is appropriate for sufficiency; that required for decent living but no more. Rao and Min35 offer the basis from which the values in the present work (and previous DLE model) are derived, and how these are translated to be suitable for an energy model is described fully in previous work22. Estimating energy intensities appropriate for state-of-the-art technologies as we intend requires harvesting and assimilating data from a broad range of sources including life cycle assessment, input–output analysis and industrial ecology. This process is also fully described in previous work22. For the present model, an analogous process is undertaken to obtain energy intensities for the current technology scenario, which is described in the Supplementary Information. Note also that both activity-levels and energy intensities are made regionally-variable where appropriate, and where suitable data exists.
The nature of bottom-up models is that some of the limitations of top-down models are avoided, but others are introduced. Most prominently, when one compiles an inventory of material consumption assumed sufficient for decent living, it is far more likely that important sectors will be missed than unnecessary sectors (mistakenly) included. The present model, for example, includes government services across healthcare and education, but it doesn’t include police or military activity; it includes household consumption to support individual needs and social participation, but doesn’t explicitly include artistic or cultural activities. It thus tends towards an underestimation of energy requirements. Other limitations relate to its static nature – the focus on a single year – and consequent lack of assessment of the feasibility of state-of-the-art technologies being fully deployed by 2050, given the lifetimes and inertia of current infrastructures, not to mention social and political lock-in.
Finally, note that while global estimates of energy use are reported here, we only make estimates for 120 countries, which align with the GTAP regions used in the original DLE work22. However, these cover 89% of the 2050 global population and over 95% of current global GDP. And hence to obtain global estimates, we scale up the total energy use for the 120 countries by 112% (i.e. 1/0.89).
Modelling final energy
Final energy is modelled as this better reflects the energy requirements of society and economic activity54, as opposed to primary energy that captures losses during conversion of fossil fuels – e.g. coal into electricity, or oil into gasoline – losses that have no analogue for renewable energies. Final energy is, however, still a means to an end – specifically, to an energy service such as heating or mobility. These energy services provide benefits like comfort and social participation, which may satisfy various dimensions of human well-being. Decent living standards may thus not be broadly met even in societies with high final energy use (in excess of the DLE estimate) as energy use may be distributed highly unequally, provisioned via inefficient technologies, or directed towards energy services that are at odds with human well-being.
Modelling ‘fair’ inequalities
Social and political scientists have studied public attitudes to inequality in various ways. We draw upon data reporting people’s ‘ideal’ income ratios between the highest earners and unskilled workers for 40 countries39, of which 39 overlap with the 120 considered here (Iceland accounting for the difference). These ratios are lowest in Scandinavia and some Eastern European countries (at 2-3), higher in Germany and the USA (~7) and highest in Taiwan and South Korea (>10). Other important conclusions from this field of research are that notions of fair inequality are surprisingly consistent across countries, socioeconomic status, and political identities39, and that almost all data suggests people significantly underestimate the extent of current inequalities55.
We follow Millward-Hopkins and Oswald (2021)24 to convert these maximum income ratios into idealised distributions considered to describe public notions of fair inequality. The first stage involves simplifying the approach by categorising countries as egalitarians, moderates or meritocrats, depending upon the level of inequality considered fair: egalitarians are countries where the reported ideal income ratio is under 4, moderates where it’s 4-6, and meritocrats where it exceeds 6. We then produce idealised (lognormal) distributions for each group, at a resolution of deciles up to the top decile, which is split into the 90-95th, 95-99th, and top 1%. Again following Millward-Hopkins and Oswald, for the three fair distributions, ratios between the top 1% and bottom 10% are set to 2.5, 5 and 8 for egalitarians, moderates and meritocrats, respectively, leading to the distributions shown in Supplementary Figure 1 (see Supplementary Information). For the Fairly-large inequality scenario, these distributions are widened until the GINI coefficients of the distributions become 0.25, 0.35 and 0.45, respectively (up from 0.12, 0.21 & 0.26 in the fair inequality scenario), which together roughly span the range of national income GINI coefficients currently observed (see data.worldbank.org/indicator/SI.POV.GINI).
Sensitivity analysis shows our main results to hold even when these idealised distributions are parameterised differently, consistent with Millward-Hopkins and Oswald (2021). This leaves the major limitation being the amount of missing data – for 81/120 counties, largely in Africa and Asia – and we simply categorise these countries as moderates. Note, however, that the 39 countries for which we have data cover all six continents and include the world’s major economies (e.g. China, the USA & most of Europe).
Implementing inequalities in material consumption
In Millward-Hopkins and Oswald (2021), these idealised distributions were taken as income distributions, then translated into expenditure distributions and onto carbon and energy footprints using input-output data. For the present model, however, these distributions must instead be translated directly into material consumption.
To this end, the idealised distributions are taken as dimensionless descriptions of relative consumption. The distributions are thus applied linearly to private luxuries – housing size, car travel, air travel, hot water for bathing, energy-intensive foods, etc. – while retaining decent living standards as a floor on consumption for the lowest consumers. For housing, for example, the bottom 10% have 15 m2/cap of floor space, while the top 1% have 37.5 m2 and 120 m2 in egalitarian and meritocratic countries, respectively (i.e. 15×2.5 and 15×8). Mobility is more involved, as increases in private road transport are assumed to displace public surface transport before total mobility increases (see the Supplementary Information for details, and Supplementary Figure 2 for an example). To ensure this linear scaling didn’t result in unrealistic values – there’s only so many flights even the richest may take each year, for example – a sense check was done on the resulting consumption, and limits defined based upon the maximum expected even for the wealthiest (see Supplementary Table 3). For hot water, for example, a limit of 300 L/cap/day was applied, based upon flowrates of modern luxury showerheads. As mentioned above, for the super-rich scenario consumption of the top 1% was increased further for housing and mobility, to levels based upon those reported by Otto et al.56 (also detailed in Supplementary Table 3).
The major assumption thus underpinning this process is that the income inequalities people believe to be fair can be taken to describe the inequalities in material consumption people think fair. Clearly this assumption can be challenged. However, for the present modelling approach – which is absent of monetary values – this is the only viable option, and it can be argued a reasonable approximation, as biases pull in both directions: On the one hand, applying inequalities only to a subset of the dimensions within the DLE consumption basket of Table 1, and not considering how wealthier classes will consume other luxury goods, biases the model towards underestimating the material inequalities that accompany income inequalities. Similarly, some of the things assumed to be equally distributed in the inequality scenarios are not so in reality – wealthier classes may draw more upon educational and healthcare services, for example, with children attending schools with smaller classes, and more frequent use of medical care, directed not just towards health issues but also improvements (e.g. cosmetic surgery). On the other hand, however, income inequalities frequently manifest in ways that don’t require additional material consumption – for example, expensive houses are of course not more expensive merely because they’re larger, but due also to more exclusive locations. Such factors bias the present model towards overestimates, as a proportion of income inequality will not manifest anywhere in the material consumption that the model considers.