Energy Poverty and Just Transformation in Greece

The European Commission has designed advanced and ambitious policies to tackle energy poverty, committing to protect vulnerable households. In the 2019 “Clean Energy for All Europeans” package, energy poverty was set as one of the key energy policy priorities. Although each Member State is allowed to establish its own criteria for measuring and assessing energy poverty (Faiella and Lavecchia 2021), the need to use a common definition across Europe is well-established, to further facilitate the monitoring and mitigation of the situation—and so is the need to set clear objectives for energy poverty reduction in the National Energy and Climate Plans (NECPs) of each Member State (European Commission 2019). In addition, provisions for dealing with energy poverty have been introduced in a series of European directives.

Indicatively, Directive 2019/944/EU¹ outlines the internal electricity market measures aiming at protecting vulnerable and energy-poor consumers in the context of the internal electricity market (social or energy policy measures to help pay electricity bills), launching investments to improve energy efficiency and/or protect consumers (including prohibition of disconnection of energy-poor consumers in critical periods), as well as providing measures other than public interventions in setting the prices for the supply of electricity. For the internal gas market, following up on Directive 2009/73/EE,² Directive 2019/692/EE³ details provisions for protecting energy-vulnerable households and prohibiting their disconnection in critical periods. Targeting energy efficiency, Directive 2018/2002/EC⁴ outlines measures aimed at tackling energy poverty and lowering the vulnerability of consumers; while Directive 2018/844/EU,⁵ outlining the building energy performance framework and the long-term renovation strategy of the built environment, explicitly includes actions and measures to alleviate energy poverty among European households, according to criteria established by the Member States. Finally, Regulation 2018/1999/EU⁶ (Governance of the Energy Union and Climate Action) requires that consolidated NECPs of EU Member States include estimates of the number of energy-poor households, set a national target to mitigate energy poverty, and provide progress reports including the number of energy-poor households and the policy measures in place to address the issue.

It is, therefore, evident that energy poverty is high on the European Commission’s agenda, especially amidst an energy crisis that is partly driven by Covid-19 recovery efforts as well as the skyrocketing gas and electricity prices caused by changing demand–supply dynamics and by Russia’s invasion of Ukraine in 2022. The EU policy response included the REPowerEU strategy and national measures to reduce reliance on imported gas from Russia and protect vulnerable households from the rising energy prices. It also advocated for the adoption of behavioural changes that can boost energy efficiency. In particular, the proposed measures in the REPowerEU Plan include increasing the energy savings target (from 9 to 13% reduction in final energy consumption compared to the Reference 2020 Scenario), the diversification of the energy supply mix alongside the groundwork for a new “joint purchasing mechanism” for gas imports, accelerated roll-out of renewable energy to replace fossil fuels (upgrading the 2030 target from 40 to 45%), and reduction of fossil-fuel use in all sectors. To support the implementation of the REPowerEU Plan, additional smart investments are deemed necessary across all sectors and levels (national, cross-border and EU). In essence, REPowerEU means that Member States strengthen any national efforts and proactive measures they had kicked off in as early as September 2021 to mitigate the impacts of the looming energy crisis (Sgaravatti et al. 2021).

The Greek energy policy framework comprises laws and regulations that are largely harmonised with most of the relevant EC directives and policy measures aiming at protecting vulnerable consumers and especially low-income households. However, much like delays in harmonising national and European energy efficiency legislation in Greece (Nikas et al. 2019), important EU directives (D2019/944/EU¹, D2018/2002/EU⁴) have not yet been fully adopted in the national policy framework.

Nevertheless, significant progress has been made in defining the categories of vulnerable consumers (financially vulnerable, dependent on continuous and uninterrupted energy supply, elderly, people with health problems and residents of disadvantaged areas), whereas several policy measures have been enforced to protect vulnerable consumers, including:

In addition, criteria, and procedures for the inclusion of consumers in the above measures as well as obligations of energy providers/suppliers are defined. However, since an official definition for energy poverty at national level has not yet been established (Arsenopoulos et al. 2020) and, considering the lack of a common definition across the EU, identifying the energy-poor population in the country can be challenging.

In 2021, the Greek parliament approved the National Energy Poverty Alleviation Plan (NEPAP 2021), according to the NECP (2019) provision. The plan constitutes the Greek national strategy against energy poverty for 2021–2030 and aims to outline a comprehensive understanding of the situation by mapping and analysing the characteristics of the affected households, focused on those with the highest vulnerability. The NEPAP also proposes effective planning and implementation of the necessary policy measures to achieve the quantitative goals set within the framework of the Greek NECP for a reduction by at least 50% of its relevant energy poverty indicators by 2025, and by 75% by 2030, compared to 2016 (baseline year). Specifically, the Action Plan has been based on the identification of households affected by energy poverty using specific quantitative criteria; the development of a specialised process for recording, monitoring and evaluating the course of alleviating the phenomenon until 2030; the formulation of a well-defined set of policy measures to tackle energy poverty; the development of a mechanism for monitoring and evaluating the effects of each policy measure, to assess their effectiveness or need for adjustments; and the exploration of specific schemes to address energy poverty in vulnerable households either through existing policy measures or through new ones.

There are three prevailing measurement methods to identify energy poverty, including (i) expenditure-based indicators that estimate the magnitude of energy poverty on a household by considering the household’s energy costs and income and comparing them to a selected threshold; (ii) consensual-based indicators, based on which inhabitants assess their household’s living conditions, regarding thermal comfort and other conditions (humidity, insulation, etc.), and their ability to afford expenditures required to secure healthy living conditions; and (iii) direct measurement, which sets a standard for an offered energy service (heating/cooling) and assesses energy poverty against this standard (Siksnelyte-Butkiene et al. 2021). However, Thomson and Snell (2016) suggest another measurement method to identify energy-poor households at local level, using welfare benefits, area-based approaches, demographic criteria or a mix thereof.

The first method used to measure energy poverty was the ten-percent rule proposed by Boardman in 1991. The ten-percent rule identifies as energy poor those households that spend on energy expenses more than 10% of their net income. The rule has been contested about its success to identify the phenomenon when other factors (e.g., energy efficiency, social factors, etc.) are considered. Other indices proposed to face the inadequacies of this rule include the Low Income-High Cost (Hills 2012) and the Minimum Income Standard indicator (Moore 2012). However, these are based on inflexible thresholds that are mostly theoretical, disregard actual energy costs and the equivalised ratio between income and energy-related expenditures or the difficulties in accounting different energy services (Tirado Herrero 2017).

Another approach on measuring energy poverty is that of Equivalisation of Modelled Energy Costs, proposed by Antepara et al. (2020) to face the problem of unreliable energy consumption data or behavioural practices due to socio-economic factors that may modify domestic energy use patterns. This methodology models energy bills on building energy conditions and consumption patterns and relates them to socio-demographic weighted (equivalised) variables.

To integrate more aspects, several multidimensional energy poverty indices have been established in the literature (e.g., Nussbaumer et al. 2012; Bersisa 2019; Ntaintasis et al. 2019; Crentsil et al. 2019). One of the first attempts to introduce composite indicators was that from Healy and Clinch (2002), who used 6 subjective indicators. Recently, Delugas and Brau (2018) identified energy poverty as a factor of wellbeing, Gouveia et al. (2019) combined energy performance with social and economic indicators, while Sokolowski et al. (2020) combined five quantitative (monetary) and qualitative (non-monetary) indicators, aiming to address limitations of previous single- or multi-dimensional indices.

To measure energy poverty at the European level, the EU Energy Poverty Advisory Hub (EPAH) has proposed several different indicators, four primary and nineteen secondary.⁷ The primary indicators comprise the shares of the (sub-) population that delay paying utility bills and that cannot keep their home sufficiently warm, calculated on the basis of answers to closed-ended questions (EUSilc); as well as household income and energy expenditure (with data from Household Budget Surveys—HBS)—e.g., energy expenses being more than twice the national average household income. However, caution is advised regarding structural differences in energy expenditure among households, situations where energy is often—albeit not exclusively—included in rent, and high energy efficiency standards or considerable underconsumption of energy. The secondary indicators include average prices paid by a household per kWh from district heating or generated from specific components such as fuel oil, biomass, and coal; electricity and gas prices for different types of consumers; energy consumption expenditure on electricity, gas, and other fuels as a share of income for 5 income quintiles; accommodation-related indicators such as average number of rooms per person in owned, rented, and all dwellings; location (densely populated or intermediate residential area); dwelling condition (leakage, dampness, or rot); and people at risk of poverty or social exclusion or death in winter.

The 2030 framework of the Covenant of Mayors also commits to contributing to energy poverty alleviation,⁸ based on six classes of indicators: climate, facilities/housing, mobility, socio-economic aspects, policy and regulatory framework and participation/awareness-raising. The climate category includes the energy consumption for space heating and cooling a building and the frequency of hot and cold waves per month each year. Facilities or housing indicators leverage various data that may be available at the municipal level—e.g., on buildings, households, etc. This data may include energy consumption such as the percentage of municipal energy consumption per capita over national energy consumption per capita, or information on the living conditions such as the percentage of population/households with leakage, dampness, or rot in their dwelling, the percentages of households/individuals experiencing discomfort in heating or cooling and those connected to the electricity or gas grid. Indicators also report systems for heating and cooling, e.g., central heating, oil boilers, wood boilers, conventional gas boilers and central cooling systems. There also exist indicators relating to the energy efficiency of buildings, specifically for categories F, G, H, the percentages of buildings refurbished each year that have an Energy Performance Certificate (EPC) higher than B. In addition, there are assessments on occupants, social housing flats to total flats, and social housing energy demand of the national median demand. Finally, this category is completed by indicators relating to the share of households with absolute energy expenditure above a defined share of the national average, considering the average age of buildings by period of construction, and the ratio of households/individuals relying primarily on clean fuels and related technology. The mobility category includes indicators focusing on public transport (distance to nearest station, frequency of transit to serve the public, household access to essential services by foot, bicycle, or public transport, share of social housing households without easy access to public transport and of households receiving support to pay for public transport services). Socio-economic indicators include unemployment, inability to keep the house sufficiently warm or cool, money spent on electricity or gas consumption, citizens’ spending on energy services compared to their income and national average, arrears on utility bills, age, education level, household situation (vulnerability, poverty and at-risk-of-poverty, state aid, and GDP- or income-based indicators). Policy and regulatory indicators report on the status of energy poverty strategy, incentives for owner schemes, rent regulation and specific energy-poverty policy measures. Finally, participation/awareness-raising refers to the existence of a deterrent to rent increases due to energy retrofits, and engagement and cooperation with local actors on energy poverty.

Among the several methodological approaches for identifying energy-poor households, the Hellenic Ministry of Environment and Energy has selected the 4 EPAH indicators to make an initial assessment of the situation in the country (NEPAP 2021). The Centre of Renewable Energy Sources and Saving (CRES), in charge of the National Observatory of Energy Poverty, has proposed an additional indicator, namely the coverage of basic energy needs per household, calculated as the ratio of actual recorded energy consumption to theoretically required energy consumption for specific uses; particular attention is paid at how required household energy uses are determined.

In the context of the NEPAP (2021), it is desirable to identify energy-poor households through a combined multi-dimensional index, which considers as many factors as possible, as proposed by Directive 2019/944/EU¹ (and related to income, purchasing costs of energy products, and energy efficiency of residential buildings). Finally, an additional indicator, I&II_eq, is also included to target the households that simultaneously meet both of the following requirements:

According to the I&II_eq indicator’s initial calculation, the percentage of affected households in 2016 amounted to 14% of all households (approximately 573,000 households). The latest progress report,⁹ issued in 2021, calculates the I&II_eq at about 12% (approximately 497 thousand households). Continuous evaluation of this indicator to measure energy poverty is a priority, in order to make adjustments to policy measures and regulations as needed according to new scientific evidence on the factors that impact energy poverty.

Rising income inequality is a global concern, implying that economic growth is not inclusive and its benefits are not equally distributed to all households (EC 2017). Income inequality can reduce economic growth, while raising concerns about sustainable growth, as the gap between rich and poor widens (EC 2018b). Income inequality is defined as inequality in earnings received from employment, private income from investments and property, transfers between households, state benefits, pensions and rent (UN 2015).

There has been considerable debate on the drivers of income inequality (IMF 2015), which typically include:

Most of these drivers are featured in GEM-E3-FIT modelling framework, including:

Table 2 includes the most common indicators to measure income inequality. The Gini coefficient is the most established and popular indicator, while the decile dispersion ratio presents the ratio of the average income of two deciles. However, this indicator does not use information about the distribution of income within deciles and does not provide information about incomes in the middle of the distribution. Other indicators have been developed to improve understanding about income distribution, e.g., the Generalised Entropy family (e.g., the Theil index) and the Atkinson index.

The analysis of distributional impacts focuses on the income changes between deciles. The indicators of income inequality can be estimated using a combination of modelling results, and additional income data. This chapter focuses on the Gini coefficient and the Decile Dispersion ratio (S80/S20), as these indicators complement each other.¹⁰

Extreme poverty and income inequality have decreased globally after 2000, largely owing to the decrease in inequality between countries (Revenga and Dooley 2019). In 2020, the Gini coefficient for the EU was 30.0, compared to 30.2 in 2010, showcasing stability in income inequality. Bulgaria has the highest Gini coefficient, while the lowest inequality in the EU is observed in Belgium, Finland, Czech Republic, Slovenia and Slovakia (Eurostat 2019). Greece registers a value close to the European average, with signs of inequality reduction in the last decade (the Gini coefficient declined from 32.9 in 2010 to 31.4 in 2020). In 2020, the EU-28 S80/S20 ratio was 4.9, implying that the richest 20% of the population receives about five times higher income relative to the poorest 20%. This share has been stable in the 2010s. Greece ranks slightly higher than the EU average, but with clear signs of reducing inequalities, with the S80/S20 indicator declining from 5.61 in 2010 to 5.23 in 2020.

A combination of energy-inefficient housing and appliances, high energy prices and low-income levels typically determine if a household is at risk of energy poverty (Pye et al. 2015; Gouveia et al. 2019). Our study focuses on expenditure-based indicators to assess energy poverty dynamics that can be quantified using GEM-E3-FIT model outcomes on energy expenditure and income per decile, in particular the share of energy expenditure in income (2 M). This indicator measures the share of households, whose share of energy expenditure relative to their disposable income is more than twice the national median share. The highest income group has a very low share of households in energy poverty (Bouzarovski et al. 2020, as the richer a household is, the lower the share of income is dedicated to energy expenditure. The proportion of households whose share of energy expenditure in income is more than twice the national median share is estimated at 16.2% in 2015 in the EU, with Greece being very close to the EU average (Fig. 1). However, this hides remarkable differences across income deciles in Greece, with more than half of those in the first decile facing substantial energy poverty risks.¹¹

A bar graph estimates the share of energy expenditure in income by income decile in Greece. The values are as follows. Average, 17. Decile 1, 55. Decile 2, 30. Decile 3, 22. Decile 4, 18. Decile 5, 14.5. Decile 6, 12. Decile 7, 7. Decile 8, 3. Decile 9, 2.5. Decile 10, 1. Values are estimated.

Macroeconomic models enhanced with a representation of different socio-economic groups (e.g., income classes) can be used to evaluate the distributional implications of climate policies. Despite the challenges in terms of data integration and computational modelling issues, the introduction of multiple households enhances the capability of conventional macroeconomic models to assess income distribution effects (Zhang 2019). The representation of multiple households in CGE modelling has been long established (Cockburn 2001; Rutherford et al. 2005; Balasko and Tourinho 2014) but is usually constrained by limited data availability. There are ways to differentiate households, but income class is the most relevant for distributional analysis. The main caveat of this approach is that it does not capture inequality within the income deciles and the fact that households can switch deciles and change compositions (CPB 2011).

GEM-E3-FIT is multiregional, multi-sectoral, recursive-dynamic, providing details on the macroeconomy and its complex interactions with the environment and the energy system. The model has been recently enhanced with a representation of ten income classes aiming to assess the distributional implications of climate policies. It simultaneously represents 46 regions (including the EU countries individually) and 53 activities linked through bilateral trade flows and runs until 2050 (E3Modelling 2017). It covers the interlinkages between productive sectors, consumption, labour and capital, bilateral trade, investment dynamics and price formation of commodities. GEM-E3-FIT formulates the supply and demand behaviour of economic agents that are assumed to exhibit optimising behaviour while market-derived prices are adjusted to clear markets. It allows for a consistent comparative analysis of policy scenarios as it ensures that the economic system remains in general equilibrium.

Industries operate within a perfect competition market regime and maximise profits, considering the possibilities of input substitutions between capital, labour, energy and materials. Household demand, savings and labour supply are derived from utility maximisation using a linear expenditure system (LES) formulation. Households receive income from labour supply and from holding shares in companies. Investment by sector is dynamic, depending on adaptive anticipation of capital return and activity growth by sector. A distinctive feature of GEM-E3-FIT (Fig. 2) is the representation of imperfect labour markets through involuntary unemployment, simulated by an empirical labour supply equation that links wages and unemployment levels for five labour skills.

A block diagram explains the GEM-E 3-FIT model. It consists of households, labor supply, consumption, government, bank, total production, domestic use, composite good, capital demand, intermediate demand, labor demand, and investment.

Various policy instruments can be represented in GEM-E3-FIT, including energy and climate measures, and their interactions with policies related to labour market, economy, trade and innovation. Policies are evaluated based on their impact on sectoral growth, income distribution, employment, economic competitiveness and GDP. GEM-E3-FIT can assess the impacts of market-oriented policy instruments, such as carbon taxes and pollution permits, and investigate market-driven structural changes. It can analyse policy impacts in the allocation of capital, income, trade and labour, and provide insights on compensating measures aiming at alleviating adverse distributional effects among and within countries.

The representation of multiple households in CGE models is a challenging task both from a computational and data point of view (Zhang 2019). In the current study, this is implemented with a linkage of the CGE GEM-E3-FIT model with a satellite module with multiple households, through a sequential exchange of prices, incomes and demands until an equilibrium point is established (Rutherford and Tarr 2004; Rausch et al. 2011). This approach is easier to implement in large-scale CGE models with manageable computational complexity (compared to the hard-link approach) and it is thus adopted in GEM-E3-FIT. This was driven by the empirical findings of Rutherford and Tarr (2004), who showed very limited benefits from using the hard-link representation compared to the sequential approach. Considering the large geographic and sectoral granularity of the model and the need for short running time, we adopted the soft-link approach in the current study. The methodology is described in detail by Fragkos et al. (2021).

The equivalised disposable income by decile is used to calculate inequality indicators, but GEM-E3-FIT produces total disposable income by decile. The estimation of equivalised disposable income can be implemented by assuming that the equivalised household size by decile remains constant over 2015–2050 combined with a simplified assumption about the number of households by decile. GEM-E3-FIT results on income by decile can be used to quantify the Gini coefficient, by estimating 10 points of the Lorenz curve, each one representing an income decile group. The area under the Lorenz curve can be calculated by summing the areas of the 10 trapeziums, allowing to estimate the Gini coefficient as equal to the area below the line of perfect equality minus the area calculated below the Lorenz curve divided by the area below the line of perfect equality (0.5). The decile dispersion indicators (S80/S20 ratio) can be directly estimated via the income by decile quantified by GEM-E3-FIT. The disposable income of the last two and first two decile groups is utilised to calculate the S80/S20 indicator using GEM-E3-FIT results.

The “share of energy expenditure in income” indicator is used to identify energy poverty and can be estimated using the decile’s total energy expenditure and total disposable income. However, this indicator requires information on the distribution of absolute energy expenditures and incomes on household level to derive changing median values. GEM-E3-FIT cannot provide the median of these indicators; thus, we estimate only an adjusted 2 M indicator using the model output (i.e., the average share of energy expenditure in income by decile). While this approach does not reflect the unequal distributions within a decile, it provides insights regarding the economic burden of energy-related expenses on household budgets. In the analysis below, two indicators are used to measure the impacts of decarbonisation by decile group, namely “the share of energy expenditure for fuels and electricity in household income” (Indicator 1) and “the share of energy expenditure for fuels/electricity and energy equipment in household income (Indicator 2)”.

The integration of multiple households in GEM-E3-FIT requires data for disaggregating household expenditure by product category and income class, household earnings by branch and income class and data for calculating energy poverty indicators. Two key data sources are used: the EU Survey on Income and Living Conditions (SILC) and the Household Budget Survey (HBS).¹² We use SILC data for disaggregating income sources and HBS data for detailing household consumption and calculating expenditure-based energy poverty indicators. Income and expenditure disaggregation are based on the micro data on an average per-household level and per-income decile. Data for each country and for each income decile is extracted from the SILC and HBS microdata for the latest available year, including data for the structure of population (e.g., number and size of households, occupation), income (income sources per occupation, benefits, transfers, allowances, dividends and property income and saving rates), expenditures (taxes, transfers and consumption per purpose) and indicators on energy poverty and income inequality. It should be noted that the model-based development and data analysis is conducted at muti-regional pan-European level (with a focus on Greece) as other European countries influence the national Greek economy through various modelling channels, including international trade, technology innovation capital and labour transfers.

Income deciles are constructed using national household sample weights included in the HBS dataset and each household in the dataset is assigned to a decile. Subsequently, average expenditures of households within each decile by Classification of Individual Consumption by Purpose (COICOP) category are calculated, using sample weights to obtain the actual distribution in the population. For energy expenditure and net income, standard deviations and skewness by income decile are calculated in the same way. To calculate the average share of energy expenditure in household income by decile, household energy expenditures are divided by incomes. Then, weighted averages, standard deviations and skewness by country and income decile were calculated, using household sample weights. Data for other variables are extracted from the SILC database: total gross and disposable incomes, decile-specific top cut-off points, standard deviations, skewness, income per occupation (ISCO-88), tax payments, income from other sources: various household and personal-level benefits and transfers, interests, properties and pensions.

Two key data-related challenges have emerged: the first relates to data structure that represents how individuals are nested in households, and the second to GEM-E3-FIT representing only household income. As household-level information is key to macroeconomic modelling, we assigned household-level income deciles to individuals for calculating resulting per-capita averages and per-decile totals. For most variables, macro-level data published by Eurostat cannot be replicated and thus Eurostat data is used for calculating the respective shares.

The Reference (REF) scenario is a projection of the future evolution of the global economic and energy system based on existing trends, exogenous assumptions and scientific expertise on specific fields. Socio-economic developments replicate IEA assumptions (IEA 2019) and are consistent with the SSP2 scenario widely used by the IPCC. For the EU, socio-economic assumptions are based on the Ageing Report of the European Commission (EC 2018a, 2018b). The scenario assumes that already adopted climate policies and pledges, including the Nationally Determined Contributions (NDCs), are implemented by 2030. After 2030, no additional emission reduction effort is assumed, implying that the carbon prices resulting from NDCs in 2030 are kept constant until 2050. The costs of power generation and other energy technologies are calibrated to IRENA (2020), while technology progress is included for low-carbon technologies. ETS carbon revenues are recycled through the public budget.

We also develop a scenario consistent with the Paris Agreement goal to limit global temperature increase to well below 2 °C, which is proxied with the imposition of a global cumulative CO₂ budget of 1000 GtCO₂ over 2010 in line with the IPCC 6^th Assessment Report (IPCC 2022). A universal carbon price is implemented from 2020 onwards to reach the global cumulative CO₂ budget. As the stringency of the mitigation effort increases over time, the global carbon tax grows from 80$/tnCO₂ in 2030 (in line with IEA 2019) to about 350$/tnCO₂ in 2050. In addition, the EU implements the Green Deal targets of GHG emission reduction of 55% in 2030 and net-zero transition by mid-century. The policy mix adopted to drive the EU energy system decarbonisation includes various instruments—e.g., strengthened EU ETS, subsidies insulation in buildings, accelerated expansion of renewable energy, ambitious technology standards, increased electrification of energy services and uptake of innovative mitigation options (e.g., carbon capture storage, hydrogen, etc.) (Table 3).

In the REF scenario, economic activity and emissions are found to gradually decouple by 2050 with emissions intensity of GDP declining in all countries, by about 2% annually over 2020–2050. This is a result of the accelerated uptake of renewable energy and low-carbon technologies, the more efficient use of energy resources, fuel switching and stricter environmental regulations. As we aim to analyse the distributional impacts of climate policies, the disaggregated household-related projections by income decile should be constructed to ensure consistency with the national-level GEM-E3-FIT outputs—here, for Greece. The projections for income deciles are based on the income level and source, occupation, consumption patterns and savings, through empirically estimated relations to disaggregate household energy demand by income decile.

In Greece, the recent evolution of income distribution among deciles shows limited changes in the last decade with small variations in the income shares of the bottom and top income quintiles. As there are large differences in earnings of different occupation types, the income distribution in the study is modelled via wage evolution of different occupation types and their respective distribution across deciles depending on sectoral evolution and labour intensity of the economy. According to OECD (2006), savings are highly concentrated at the top of the income distribution and saving rates increase with income. Over time, saving rates by decile as a percentage of disposable income do not vary largely (Eurostat 2019) and can be assumed as constant in the REF scenario by 2050. The disaggregation of GEM-E3-FIT output into income deciles requires additional assumptions:

Technical progress, ageing population, changes in consumer behaviour, consumption patterns, industrial competitiveness and policies shape the structure of socio-economic developments in Greece. Changes in income distribution in the country are largely driven by GDP growth, labour supply and demand, technical progress, sectoral growth, wage differentials across skills, the distribution of skills, capital earnings and transfers across deciles. The composition of value added differs significantly across sectors indicating that policies can have different distributional effects across countries. Income inequality is mostly influenced by inequality in labour skills and wages (Keeley 2015; Harrison et al. 2011). In the model, each household receives a share of the total wage income, based on the distribution of income by decile for each skill. The wage income from low-skilled occupations and service workers is more equally spread across different deciles, while income from high-skilled occupations (e.g., managers, technicians) and dividends is mostly directed to higher income deciles in Greece. In contrast, low-income households receive most of social benefits and other allowances. Economic development in the REF scenario involves a gradual transition towards a more service-oriented, technology-rich economy, resulting in a slight redirection of labour demand towards higher skills. Over 2020–2050 there is a slight decline in the value-added share generated by low-skilled occupations (e.g., agricultural jobs) and an increase in the share of higher skilled jobs (e.g., managers). The change in occupations and labour skills has direct impacts on the distribution of income (Fig. 3). As demand for high skills grows, a higher share of total wages would be directed towards higher deciles. Low-income groups receive income mainly from low-skill occupations, and thus their income is negatively affected. However, low-income deciles are dependent on government benefits and allowances, while higher deciles receive income mostly from labour and capital endowments.

A stacked bar graph plots the composition of Greek labor. The values are as follows. 2020. Managers, 32.8. Clerks, 13.5. Technicians, 24.2. Service workers, 12.3. Low-skilled, 17.2. 2050. Managers, 33.3. Clerks, 13.8. Technicians, 24.2. Service workers, 12.7. Low-skilled, 16.

The transition towards a high-skilled and capital-intensive Greek economy results in increasing income inequality in the REF scenario as indicated by the Gini coefficient and the S80/S20 index. The inequality indicators strongly depend on the assumption of constant distribution of occupations across deciles, which faces limitations as it does not consider the impacts of potential labour supply adjustments induced by the transition to high skills, which may change skill distributions across the deciles.

The REF scenario dynamics result in limited increase in the Gini coefficient in Greece from 2020 levels (30.5%) by 0.7 percentage points (pp) in 2030 (31.2%); however, in the longer term, increased automation, digitisation and higher requirements for labour skills drive a larger increase in income inequality, with the Gini coefficient increasing to 33.8 in 2050 (Fig. 4). Greece is expected to remain close to the EU average in terms of the Gini coefficient until 2050 (the EU Gini coefficient is projected to slightly increase from 30% in 2020 to 32.8% in 2050). The S80/S20 indicator is projected to increase over time in the REF scenario, from 5.2 in 2020 to 5.45 in 2030, and further to 6.1 in 2050.

A bar graph estimates the evolution of the Gini coefficient in Greece in the R E F scenario. The values are as follows. 2020, 30. 2030, 31. 2040, 32.9. 2050, 34. Values are estimated.

The implementation of ambitious climate policies towards the long-term climate neutrality target (DECARB scenario) would have large-scale impacts triggered by the accelerated uptake of renewable energy and energy efficiency, the massive electrification of end-uses and the deployment of CCS and green hydrogen. The imposition of high carbon pricing drives energy system transformation towards a more capital-intensive structure, with increased investment to renewable energy, electric vehicles and energy efficiency projects, leading to increases in CAPEX and a drop in OPEX and energy purchasing costs. As GEM-E3-FIT assumes optimal use of available capital resources in the REF scenario, reallocation of investments towards low-carbon, energy-efficient technologies in the DECARB scenario leads to the so-called “crowding-out effect”, as firms and households finance their clean energy investment by spending less on other commodities and investment purposes. High carbon prices increase the cost of energy services for firms and households and, hence, production costs throughout the economy, with a depressing effect on consumption and GDP; this is partly alleviated by increased investment in low-carbon technologies. Overall, the net-zero transition is projected to lead to a slowdown of EU economic growth by 0.3% in 2030 and 1.1% in 2050 compared to the REF levels with differential impacts across countries, depending on their economic structure, their relative position in international trade (especially for fossil fuels and low-carbon technologies) and the mitigation effort. The DECARB scenario impacts differently specific sectors in Greece, with sectors directly related to fossil fuels (e.g., mining, refineries, and fossil-based power plants) facing pronounced negative effects due to the shift towards low-carbon energy sources. In contrast, increased electrification of energy services drives increased activity in the electricity sector, required for electric vehicles and heat pumps, and the emergence of green hydrogen after 2035. The output of energy-intensive industrial sectors declines by about 2% with regard to the REF scenario, due to their carbon-intensive structure, as energy costs represent a high share of production costs. Energy efficiency improvements imply increased requirements for construction directed to building retrofits and thermal insulation.

The employment impacts of DECARB are relatively limited and are driven by declining GDP counterbalanced by the uptake of more labour-intensive technologies—e.g., renewable energy and energy efficiency (Fragkos and Paroussos 2018). This aggregate effect masks large differences across productive activities, with some sectors facing extensive job losses (e.g., lignite mining, refineries) due to reduced output, while others are influenced positively by the transition (e.g., electricity, construction). These employment shifts across sectors require extensive re-allocation of workforce and the development of labour skills related to decarbonisation. The labour markets will be influenced by the transition, not only in activities directly linked to the transition, but also for workers at various levels of the supply chain or in sectors that observe a knock-on impact through multiplier effects (construction, agriculture).

The DECARB scenario would have a depressing effect in the Greek economy with GDP projected to decline by about 1% in 2050 compared to REF levels. Private consumption and employment are also negatively impacted. Higher unemployment levels would negatively influence average wage rates, with total income reducing by 0.4% in 2030 and 1.5% in 2050. The largest impacts are felt in low-income deciles, with the income of the lowest decile dropping by 2% from REF levels in 2050, while impacts are limited (less than 1%) in high-income groups; overall, we project a slightly increasing income inequality in Greece. This implies a slight increase in Gini coefficient from 33.8 in REF to 34.2 in DECARB in 2050.

The net-zero transition would also impact the composition of the Greek value added with increased share of high-skilled occupations to the detriment of low-skilled ones, due to higher demand for high-skilled labour required for the transition and the wage differential across different occupations and skills (as increased demand for managers results in a relatively higher increase in their respective share in income). The skills transition entails replacement of labour-intensive and low-skill occupations (Fig. 5), such as lignite mining, by skill-intensive occupations for the design, manufacturing, development and installation of clean technologies and innovative low-carbon products—these include manufacturing and software engineers, project designers, land development advisors and other high-skilled professional or managerial positions (Fragkos et al. 2021).

A stacked bar graph of the changes in the composition of skilled labor. The values are as follows. 2030. Managers, clerks, technicians, and service workers, 0.05%. Low-skilled, minus 0.05%. 2040. Managers, clerks, technicians, and service workers,1.6%. Low-skilled, minus 1.7%. Values are estimated.

The deep energy system transformation towards net-zero requires large-scale investment by households targeted to the renovation of buildings and the purchase of energy-efficient equipment and low-carbon technologies, which are capital-intensive and increase CAPEX, thereby posing challenges for low-income classes. The latter cannot afford energy-efficient appliances, houses and cars, which fuels the threat of energy poverty. Energy affordability is affected by how income inequality changes due to decarbonisation, as the amount of disposable income available for energy-related expenditure is impacted. The energy expenditure indicators (introduced in the previous section) are quantified using GEM-E3-FIT results combined with data on expenditure for fuels, electricity and energy equipment from PRIMES-Buimo (Fotiou et al. 2022).

The share of energy expenditure in income differs across income deciles, indicating the different levels of vulnerability to changes in energy prices, with lower income deciles having the highest vulnerability. The share of energy expenditure to income (Indicator 1) is estimated at 19% in 2020 for the lowest decile in Greece but is only 3% for the high-income deciles (Fig. 6). When also considering the expenditure for energy equipment (including appliances, heating devices, cars, transport equipment), the share of energy expenditure (Indicator 2) increases by an average of 5.5 pp, ranging from 8 to 22% across income deciles. The increases are relatively higher in high-income deciles that commonly purchase more expensive equipment, highly efficient appliances and more luxurious cars relative to low-income groups.

A bar graph plots the energy expenditure indicator 1 by income decile in Greece in 2020. The values are as follows. D 1, 18%. D 2, 14%. D 3, 14%. D 4, 11%. D 5, 10.5%. D 6, 9.5%. D 7, 9%. D 8, 8%. D 9, 6.5%. D 10, 3%. Values are estimated.

In the REF scenario, the share of energy expenditure to income declines somewhat in the longer term across income deciles, as household incomes increase faster than energy consumption and energy prices. The same trend is also evident when expenditure for energy and transport equipment is considered as well (Indicator 2), with this indicator dropping by about 1–2% across income groups.

Decarbonisation entails substantial changes in household energy-related expenditure, along with the subsequent income distributional changes described above. Strong carbon pricing, increased prices for energy products and the need to purchase more expensive low-carbon and efficient equipment result in increased energy expenditure across income deciles. The Energy Expenditure Indicator 1 increases by about 0.2–0.8 pp from REF in 2050, driven by increased energy-related expenditure and slight reduction in household income. The highest increases are found in low-income classes, indicating additional challenges to purchase energy and mobility services leading to higher threats for energy poverty increase. Higher increases are calculated based on the Energy Expenditure Indicator 2 that include the increased costs to purchase advanced energy-efficient equipment and low-emission cars. The highest increase is projected for low-income deciles (Fig. 7).

A stacked bar graph plots the energy expenditure indicators in Greece for each income decile in 2050 in the DECARB scenario. The stacked bars of indicator 1 R E F, indicator 2 R E F, and DECARB plot decreasing trends from 23.5% in D 1 to 7% in D 10. Values are estimated.

The reduced income and increased energy expenditure would have larger impacts on low-income groups, increasing their vulnerability to energy poverty. Several measures are discussed to alleviate such risks and pave the way for a just transition, most of them based on different ways to use carbon revenues (Budolfson et al. 2021). Here, we quantitatively assess the distributional impacts of directing the carbon revenues via lump-sum to households and via reduced social security contributions (DECARB_EQ), instead of recycling them through the public budget. The distribution of lump-sum transfers to different income groups follows the distribution of social benefits and allowances. The additional carbon revenues amount to about 1–1.5% of GDP and thus the redirection of carbon revenues has an important effect on income inequality. The available income of Greek households increases by more than 1% from REF levels, but this is moderated by the macroeconomic impacts of the transition in the longer run. Increased government benefits have large positive impacts for lower income deciles that largely depend on these benefits and allowances (their incomes increase by close to 2.5% from Reference), while the impacts are lower for high-income households and even turn negative in 2050, due to the reduced economic activity with GDP declining relative to the REF scenario. This leads to reduced income inequalities, as reflected in a reduction of the Gini coefficient, induced using carbon revenues as lump-sum transfers that mostly benefit lower income households. This measure counterbalances the regressive impacts of the skills transition and reduces income inequality with the Gini coefficient in Greece declining by 0.8 pp and 1.4 pp in 2030 and 2050, respectively, relative to REF, with similar trends observed also in the S80/S20 index (Fig. 8).

A cluster bar graph plots the difference in income by the decline from reference. 2030. The D 1 to D 10 plot a generally decreasing trend. 2040. The D 1 to D 10 plot a generally decreasing trend. 2050. The D 1 to D 10 plot a generally decreasing trend, with D 9 and D 10 plotting negative values.