Socio-economic conditions for satisfying human needs at low energy use: An international analysis of social provisioning
Received 26 July 2020, Revised 27 April 2021, Accepted 7 May 2021, Available online 29 June 2021.
Limiting global warming to 1.5 °C without relying on negative emissions technologies requires not only rapid decarbonisation of global energy systems but also deep reductions in global energy use (Grubler et al., 2018, IPCC, 2018). At the same time, billions of people around the globe are still deprived of basic needs, and current routes to sufficient need satisfaction all seem to involve highly unsustainable levels of resource use (O’Neill et al., 2018). The way societies design their economies thus seems misaligned with the twin goals of meeting everyone’s needs and remaining within planetary boundaries (O’Neill et al., 2018, Raworth, 2017). This study addresses this issue by empirically assessing how the relationship between energy use and need satisfaction varies with the configurations of key socio-economic factors, and what configurations of these factors might enable societies to meet human needs within sustainable levels of energy use.
While these questions are poorly understood and empirically understudied (Brand Correa and Steinberger, 2017, Lamb and Steinberger, 2017, O’Neill et al., 2018, Roberts et al., 2020), the corner pieces of the research puzzle are largely in place. We roughly know the maximum level of final energy use (~27 GJ/cap) that can be globally rendered ecologically ‘sustainable’ (compatible with avoiding 1.5 °C of global warming without relying on negative emissions technologies) with deep transformations of energy systems (Grubler et al., 2018, IPCC, 2018). We understand what defines and characterises human needs, and what levels of which goods, services and conditions generally satisfy these needs (Doyal and Gough, 1991, Max-Neef, 1991, Millward-Hopkins et al., 2020, Rao and Min, 2018a).
We also know the basic characteristics of the cross-country relationship between energy use and a wide range of needs satisfaction indicators, including life expectancy, mortality, nourishment, education, and access to sanitation and drinking water (Burke, 2020, Lambert et al., 2014, Mazur and Rosa, 1974, Rao et al., 2014, Steinberger and Roberts, 2010). While at low levels of energy use, these need satisfaction indicators strongly improve with increasing energy use, they generally saturate at internationally moderate levels of energy use (ibid.). Beyond that saturation level, need satisfaction improvements with additional energy use quickly diminish, reflecting the satiability of needs (Doyal and Gough, 1991).
How much energy use is required to provide sufficient need satisfaction is only scarcely researched, and the few existing estimates are broadly scattered (Rao et al., 2019). Empirical cross-national estimates include 25–40 GJ/cap primary energy use for life expectancy and literacy (Steinberger and Roberts, 2010), or 22–58 GJ/cap final energy use for life expectancy and composite basic needs access (Lamb and Rao, 2015). Empirically-driven bottom-up model studies estimate the final energy footprints of sufficient need satisfaction in India, South Africa and Brazil to range between 12 and 25 GJ/cap (Rao et al., 2019), based on Rao and Min’s (2018a) definition of ‘Decent Living Standards’ that meet human needs. Global bottom-up modelling studies involving stronger assumptions of technological efficiency and equity, respectively, suggest that by 2050, Decent Living Standards could be internationally provided with 27 GJ/cap (Grubler et al., 2018) or even just 13–18 GJ/cap final energy use (Millward-Hopkins et al., 2020). Together, these studies demonstrate that meeting everyone’s needs at sustainable levels of energy use is theoretically feasible with known technology.
What remains poorly understood, however, is how the relationship between human need satisfaction and energy use (or biophysical resource use) varies with different socio-economic factors (Lamb and Steinberger, 2017, O’Neill et al., 2018, Steinberger et al., 2020). A small number of studies offer initial insights. The environmental efficiency of life satisfaction, presented as a measure of sustainability, follows an inverted-U-shape with Gross Domestic Product (GDP), increases with trust, and decreases with income inequality (Knight and Rosa, 2011). The carbon or environmental intensities of life expectancy, understood as measures of unsustainability, increase with income inequality (Jorgenson, 2015), urbanisation (McGee et al., 2017) and world society integration (Givens, 2017). They furthermore follow a U-shape with GDP internationally (Dietz et al., 2012), though increasing with GDP in all regions but Africa (Jorgenson, 2014, Jorgenson and Givens, 2015), and show asymmetric relationships with economic growth and recession in ‘developed’ vs. ‘less developed’ countries (Greiner and McGee, 2020). Their associations with uneven trade integration and exchange vary with levels of development (Givens, 2018). Democracy is not significantly correlated with the environmental efficiency of life satisfaction (Knight and Rosa, 2011) nor with the energy intensity of life expectancy (Mayer, 2017). All of these studies either combine need satisfaction outcomes from societal activity and biophysical means to societal activity into a ratio metric, or analyse residuals from their regression. Hence, they do not specify how these socio-economic factors interact with the highly non-linear relationship between need satisfaction and biophysical resource use, or with the ability of countries to reach targets simultaneously for need satisfaction and energy (or resource) use.
The socio-economic conditions for satisfying human needs at low energy use have been highlighted as crucial areas of research (Brand Correa and Steinberger, 2017, Lamb and Steinberger, 2017, O’Neill et al., 2018, Roberts et al., 2020), but remain virtually unstudied. While the theoretical understanding of this issue has seen important advances (Bohnenberger, 2020, Hickel, 2020, Stratford, 2020, Stratford and O’Neill, 2020, Gough, 2017, Kallis et al., 2020, Parrique, 2019), empirical studies are almost entirely absent. Lamb, 2016a, Lamb, 2016b qualitatively discusses socio-economic factors in enabling low-energy (or low-carbon) development, but only for a small number of countries. Furthermore, Lamb et al. (2014) explore the cross-country relationship between life expectancy and carbon emissions in light of socio-economic drivers of emissions, but do not quantitatively assess how life expectancy is related to carbon emissions nor to socio-economic emissions drivers. Quantitative empirical cross-country analyses of the issue thus remain entirely absent.
We address these research gaps by making three contributions. First, we develop a novel analytical approach for empirically assessing the role of socio-economic factors as intermediaries moderating the relationship between energy use (as a means) and need satisfaction (as an end), thus analytically separating means, ends and intermediaries (Fig. 1). For this purpose, we adapt and operationalise a novel analytical framework proposed by O’Neill et al. (2018) which centres on provisioning systems as intermediaries between biophysical resource use and human well-being (Fig. 1A). Second, we apply this approach and framework for the first time, using data for 19 indicators and 106 countries to empirically analyse how the relationships between energy use and six dimensions of human need satisfaction vary with a range of political, economic, geographic and infrastructural ‘provisioning factors’ (Fig. 1B). Third, we assess which socio-economic conditions (i.e. which configurations of provisioning factors) might enable countries to provide sufficient need satisfaction within sustainable levels of energy use. Specifically, we address the following research questions:
- 1) What levels of energy use are associated with sufficient need satisfaction in the current international provisioning regime?
- 2) How does the relationship between energy use and human need satisfaction vary with the configurations of different provisioning factors?
- 3) Which configurations of provisioning factors are associated with socio-ecologically beneficial performance (higher achievements in, and lower energy requirements of, human need satisfaction), and which ones are associated with socio-ecologically detrimental performance (lower achievements in, and greater energy requirements of, need satisfaction)?
- 4) To what extent could countries with beneficial configurations of key provisioning factors achieve sufficient need satisfaction within sustainable levels of energy use?
The remainder of this article is structured as follows. We introduce our analytical framework and outline our analytical approach in Section 2. We describe our variables and data in Section 3, and detail our methods in Section 4. We present the results of our analysis in Section 5, and discuss them in Section 6. We summarise and conclude our analysis in Section 7.
2. Analytical framework and approach
Building on the work of O’Neill et al. (2018), our analytical framework (Fig. 1A) conceptualises the provisioning of human needs satisfaction in an Ends–Means spectrum (Daly, 1973). Our framework considers energy use as a means, and need satisfaction as an end, with provisioning factors as intermediaries that moderate the relationship between means and ends. We thus operationalise O’Neill et al.’s (2018) framework by reducing the sphere of biophysical resource use to energy use (for analytical focus), and reducing the sphere of human well-being to human need satisfaction (for analytical coherence). Our operationalisation of human need satisfaction follows Doyal and Gough’s (1991) Theory of Human Need, reflecting a eudaimonic understanding of well-being as enabled by the satisfaction of human needs, which can be evaluated based on objective measures (Brand Correa and Steinberger, 2017, Lamb and Steinberger, 2017).
The main advancement of our framework consists in operationalising the concept of provisioning systems (Brand Correa and Steinberger, 2017, Fanning et al., 2020, Lamb and Steinberger, 2017, O’Neill et al., 2018) by introducing the concept of ‘provisioning factors’.
Provisioning factors comprise all factors that characterise any element realising, or any aspect influencing, the provisioning of goods and services. This includes economic, political, institutional, infrastructural, geographic, technical, cultural and historical characteristics of provisioning systems (or the provisioning process), spanning the spheres of extraction, production, distribution, consumption and disposal. In other words, provisioning factors encompass all factors that affect how energy and resources are used to meet human needs (and other ends). For example, it matters whether provisioning caters to consumers with equal or unequal purchasing power, whether it occurs in an urban or rural context, in a growing or shrinking economy, whether electricity is available, and what transport infrastructure is in place. Provisioning factors are intermediaries that moderate the relationship between energy use and need satisfaction. Whereas provisioning systems are broad conceptual constructs that are difficult to measure, provisioning factors are tangible and measureable, and as such operational: provisioning factors characterise provisioning systems (or the provisioning process).
While interactions between energy use, provisioning factors and social outcomes may in principle go in all directions (Fanning et al., 2020, O’Neill et al., 2018), our focus here is on the role of provisioning factors for countries’ socio-ecological performance, i.e. their achievements in, and energy requirements of, human need satisfaction (Fig. 1A). We use regression-based moderation analysis (Section 4.2) to assess how the relationship between energy use and need satisfaction varies with different provisioning factors, and subsequently model that relationship for different configurations of each provisioning factor (Fig. 1B). We further estimate how multiple provisioning factors jointly interact with the relationship between need satisfaction and energy use, using multivariate regression analysis (Section 4.3). While these are established statistical techniques, the way we apply them to our analytical framework and research questions is novel. Our approach allows us to coherently assess and compare the interactions of a broad range of provisioning factors, not just with need satisfaction or its ratio with energy use, but with the relationship between need satisfaction and energy use, across the international spectrum.
The variables assessed in our analytic framework (listed in Fig. 1A and detailed in Table 1, Table 2) capture key dimensions of human need, key categories of provisioning (state provision, political economy, physical infrastructure and geography) as well as total final energy use. Based on our understanding of human need theory (Doyal and Gough, 1991, Max-Neef, 1991) and provisioning systems (Brand Correa and Steinberger, 2017, Gough, 2019, O’Neill et al., 2018, Fanning et al., 2020), we analyse electricity access, democratic quality and income equality as provisioning factors (intermediaries) rather than as indicators of human need satisfaction (outcomes).
Table 1. Human need satisfaction variables used in the analysis.
|Variable name||Description and [units]||Sufficiency threshold||Indicator source|
|Healthy life expectancy||Average healthy life expectancy at birth [years]||65 years||IHME GBD|
|Sufficient nourishment||Percentage of population meeting dietary energy requirements [%], calculated as the reverse of Prevalence of undernoursihment, rescaled onto a scale from 0 to 100%||95%||WB WDI 2020|
|Drinking water access||Percentage of population with access to improved water source [%]||95%||WB WDI 2017|
|Safe sanitation access||Percentage of population with access to improved sanitation facilities [%]||95%||WB WDI 2017|
|Basic education||Education index [score]||score of 75||UNDP HDR|
|Minimum income||Absence of income shortfall below $3.20/day [%], calculated as the reverse of the Poverty gap at $3.20 a day (2011 PPP)||95%||WB WDI 2020|
Saturation transformations are applied to all need satisfaction variables (see Supplementary Materials Section C.4.2). Indicator sources are: the Global Burden of Disease Study (IHME GBD; Institute for Health Metrics and Evaluation, 2017), the World Development Indicators (WB WDI; World Bank, 2017, World Bank, 2020), and the Human Development Report 2013 (UNDP HDR; UNDP, 2013).
Table 2. Provisioning factor variables used in the analysis.
|Variable name||Description and [units]||Trans-formation applied||Indicator source|
|Electricity access||Percentage of population with access to electricity [%]||Saturation||WB WDI 2017|
|Access to clean fuels||Percentage of population with access to non-solid fuels [%]||Saturation||WB WDI 2017|
|Trade & transport infrastructure||Quality of trade and transport-related infrastructure [score], component of the Logistics performance index||Identity||WB WDI 2017|
|Urban population||Percentage of population living in urban areas [%]||Identity||WB WDI 2017|
|Public service quality||Quality of public services, civil service, and policy implementation [score], calculated as Government effectiveness, rescaled onto a scale from 1 to 6||Identity||WB WGI|
|Public health coverage||Percentage of total health expenditure covered by government, non-governmental organisations, and social health insurance funds [%]||Identity||WB WDI 2017|
|Democratic quality||Ability to participate in selecting government, freedom of expression and association, free media [score], calculated as Voice and accountability, rescaled onto a scale from 1 to 6||Saturation||WB WGI|
|Income equality||Equality in household disposable income [score], calculated as the reverse of the Gini index||Saturation||SWIID|
|Economic growth||3-year (2010–2012) average percentage annual growth rate of GDP per capita in constant 2011 $ PPP [%], calculated based on Gujarati, 1995, pp. 169–171||Identity||WB WDI 2017|
|Extractivism||Share of total value generation obtained from total natural resource rents [% of GDP]||Logarithmic||WB WDI 2017|
|Foreign direct investments||Share of foreign direct investments (net inflow) in total value generation [% of GDP]||Logarithmic||WB WDI 2017|
|Trade penetration||Share of total value generation that is traded [% of GDP], calculated as|
3.1. Variables and data sources
We operationalise energy use in terms of total final energy use per capita, need satisfaction in terms of six key dimensions of human need (Table 1), and provisioning factors in terms of 12 diverse political, economic, geographic, and infrastructural factors (Table 2). Due to limited data availability, the assessed variables provide only a partial operationalisation of each of the three analytic domains, and are somewhat confined to variables reflecting a Western-industrial understanding of development (which have better data availability). Following O’Neill et al. (2018), we define a threshold value for ‘sufficient’ need satisfaction as a minimum societal goal for each assessed need (listed in Table 1 and discussed in Supplementary Materials Section C.1). Our energy data, sourced from the International Energy Agency (2015), provide a ‘production-based’ account of total final energy use, and hence do not account for the energy footprints of imported goods and services or international travel, due to poorer international coverage of consumption-based energy indicators. Data sources for our need satisfaction and provisioning factor variables are detailed in Table 1, Table 2, respectively.
3.2. Data sample
To ensure consistency and comparability, we use the same sample of countries throughout the analysis. Our sample, determined as the largest possible set of countries with data available for all selected variables, comprises 106 countries that together account for about 90% of the global population, 89% of global total final energy use, and 92% of global GDP. We perform a cross-sectional analysis, using 2012 as our basic year of analysis. However, we fill data gaps for 2012 in some cases by drawing on surrounding years for trade and transport infrastructure (2010–2014), income inequality (2009–2015), and minimum income (2009–2015; 2008 for Japan).
4.1. Bivariate relationship between need satisfaction and energy use
To assess the relationship between need satisfaction (NS) and energy use (ENU) across countries i, we perform bivariate linear ordinary least squares regressions, separately for each need satisfaction variable.(1)
The regression estimates the coefficient b which describes the statistical association between energy use and need satisfaction. In this case, b can be interpreted as the marginal effect of energy use on need satisfaction (mathematically:
), indicating the change in need satisfaction one would expect for a unit change in
(not necessarily a causal effect). In what follows, our use of the term ‘marginal effect’ should be interpreted in the above sense.
Throughout our analysis, all regressions are performed on transformed and standardised variables (denoted by a
). For each variable, we determine a single ‘best-suited’ transformation (Supplementary Materials Section C.4) which we use consistently throughout our analysis. On that basis, we use logarithmic transformations for our energy use variable (), and saturation transformations (as in Steinberger and Roberts, 2010) for all need satisfaction variables (), with saturation asymptotes
detailed in Table C.1 in the Supplementary Materials.
4.2. Single provisioning factors as moderators of the relationship between need satisfaction and energy use
Based on our method to determine the best-suited variable transformations (Supplementary Materials Section C.4), we apply different types of transformations (identity, logarithmic, or saturation) to different provisioning factor variables (listed in Table 2).
To assess how the relationship between need satisfaction and energy use varies with different provisioning factors, we analyse each provisioning factor separately as a moderator of the relationship between energy use and a given need satisfaction variable. In this case, moderation can be statistically estimated based on a multivariate regression of need satisfaction on energy use, a provisioning factor (PF), and their interaction term (product), as joint predictors.(2)