Assessing energy poverty indicators towards club convergence in Europe

Energy poverty, through inefficient energy utilization, can heavily impact the modern modus vivendi (i.e., way of living) with negative repercussions on human health. Energy poverty rises rapidly in tandem with the augmenting energy prices and carbon costs (Halkos & Aslanidis, 2024; Zhao et al., 2024) and can be seen as ‘’the strongest adverse social impact’’ due to its linkages to energy consumption inefficiency (Healy & Clinch, 2002), ultimately creating obstacles to energy justice and energy democracy (Bartiaux et al., 2018; Szulecki, 2018).

Urbanization has spread in Europe; henceforth climate change might also aggravate even more citizens’ lives through the urban-heat-island (UHI) effect. The UHI effect showcases that buildings’ characteristics absorb sunlight, maintaining the high morning temperatures even till thenight (Founda & Santamouris, 2017), a phenomenon that is not observed in rural areas. The UHI effect ushers citizens of greater cities to higher vulnerabilities due to climate change (Boemi et al., 2017; Curra et al., 2018), especially in the Mediterranean region due to the arid and semi-arid climate.

It has been also found that energy poverty can be attributed to excessive income spending (i.e. accessibility and affordability issues) on the coverage of energy needs, climate change implications, or even the incapacity of consuming energy due to a plethora of reasons, needs and ways (Bouzarovski & Petrova, 2015; Bouzarovski et al., 2012; Halkos & Aslanidis, 2023a; Igawa & Managi, 2022; Siksnelyte-Butkiene et al., 2021; Sokołowski et al., 2020; Song et al., 2023; Streimikiene et al., 2020). The International Energy Agency (IEA) (2023) in the “World Energy Outlook 2023” necessitated affordable transitions for households, industries, and governments, as well as secure energy transition in fuel security and trade, electricity, clean energy supply chains, and critical minerals. In total, the conservation of energy, postponing environmental degradation, and actions to lower greenhouse gas emissions can lead to sustainable development (Bampatsou & Halkos, 2018, 2019). Hence, energy poverty should be addressed in a multi-level political, societal, and economic environment through a multi-disciplinary and multi-time framework (Hasheminasab et al., 2023).

The contribution of this study is that it is not limited to the analysis of club convergence with respect to energy poverty but deepens the empirical documentation of the results of club convergence as follows: (i) It performs a robustness check for the ordered logit model, (ii) it focuses on determining the predicted probabilities of club formation based on GDP per capita and its different trends, and (iii) a test for slope homogeneity of the coefficient βi is performed. This combined approach, conducted for the first time at the level of club convergence analysis on energy poverty, provides additional information for policymakers while validating the consistency and reliability of the empirical results. The objective is to create clusters of countries that converge towards different equilibria and to identify similar or different policy approaches among countries that belong to broader geographic regions. The indicators were chosen based on the literature review for their relevance and analytical clarity. Given that the circumstances facing vulnerable households change over time, there is an urgent need for systematic monitoring of energy poverty as a basis for targeted policymaking. Europe’s unique characteristics are the social diversity, policy leadership, and data availability on energy poverty; therefore, the case study of Europe can provide valuable insights that are both region-specific and globally relevant. The motivation of the present study is to apply the log t regression test in order to answer whether there is convergence, via club clustering, among 31 European states in the period 2005–2022.

Two research questions that could be posed are (i) whether there is indeed convergence in energy poverty eradication in Europe among the 31 case studies and (ii) whether there is divergence due to geographical disparities. The structure of the study is as follows: Sect."Literature review"refers to the literature review categorized into theoretical, consensual, and empirical review; Sect."Material and methods"is attributed to the club clustering and convergence methodology; Sect."Results and discussion"presents the results, the robustness test, and the discussion for tailor-made energy poverty-oriented policies in Europe; Sect."Conclusions and policy implications"addresses the main policy implications and conclusions of the study.

An intrinsic part of environmental justice is energy justice, the importance of which rises even more in times of crises in line with other forms of vulnerabilities, inequalities, and societal-driven phobias (e.g., aporophobia, the phobia against poor people) (Bouzarovski, 2018; Grossmann & Trubina, 2021; Halkos & Aslanidis, 2023b; Nordholm & Sareen, 2021; Upham et al., 2022). Energy justice can be a three-faceted term and can be further developed into justice in the distribution of resources (i.e., distributive justice), recognition of respect and tolerance for each member in the modern fabric of society (i.e., recognition justice), and equity in decision-making (i.e., procedural justice) (Jenkins et al., 2016). Overall, the target in Europe is a more just and inclusive energy transition (Bouzarovski et al., 2021; Halkos & Aslanidis, 2023a, 2024).

In addition, the eradication of energy poverty is one of the imperatives of sustainable development goals (SDGs) by the United Nations. Energy poverty can be addressed mainly on the sub-targets of SDG 1 (i.e., eradication of poverty and all its forms) and SDG 7 (i.e., affordable and clean energy), but also in SDGs 10 (i.e., reduced inequalities) and 13 (i.e., climate action), as energy poverty can be linked with social exclusion and energy aporophobia (Halkos & Aslanidis, 2023b). Hence, energy aporophobia includes all the energy-driven and societal-focused parameters that the energy-poor citizens face; for instance, social exclusion due to inability to gather at homes and socialize, sentiments of sadness and grief due to discrimination and loss of dignity, and even higher costs, especially in rural areas or in some underserved areas in metropolitan areas (e.g., ghettos). Thus, only if energy poverty is alleviated can the achievement of Agendas 2030 (i.e., SDGs) and 2050 (i.e., Carbon-Zero Economy) be successful (Bouzarovski et al., 2021; Halkos & Aslanidis, 2023a).

Therefore, several energy poverty studies have focused on countries and regions across Europe due to the diversity in socio-economic and political structures (Thomson & Snell, 2013; Thomson et al., 2017; Herrero, 2017; Recalde et al., 2019; Karpinska & Śmiech, 2020, 2023; Stojilovska et al., 2021), by which energy poverty can escalate into a vicious circle with adverse effects on human health and security based on housing hazards and discomfort (Bienvenido-Huertas et al., 2020; Kahouli, 2020). Regarding solely the European Union (EU), parameters of energy poverty are going to be addressed in the recent multiannual financial framework (i.e., period 2021–2027) as capital will be invested in – economic, social, and territorial – cohesion (e.g., 377€ billion) (European Union, 2020). The next sub-sections are going to address the theoretical framework review upon which the energy poverty has been monitored, the review of the consensual framework for energy poverty, and the empirical review that gives novel insights into how energy poverty monitoring can be approximated through the convergence hypothesis.

The data of the present study were collected from the Energy Poverty Advisory Hub (2023) of the European Commission for 31 European countries¹ (including EU- 28 member states) and span throughout the period 2005–2022. The studied indicators are: (i) ability to pay to keep the home adequately warm, (ii) arrears on utility bills, and (iii) the presence of a leaking roof, damp walls or rotten windows, for brevity reasons these indicators would be presented as “Arrears’’, ‘’Inadequately warm’’, and ‘’Leaks’’ respectively.

The specific period includes the global financial crisis of 2008 and the starting point of the multi-crisis (that includes the energy crisis in commodity prices) in the advent of 2021. Furthermore, this specific period also covers the periods between the societal upheaval in Ukraine in 2014 as the prelude to the novel turmoil in the East. Moreover, Table 1 illustrates the descriptive statistics of the examined energy poverty indicators, which show that there is rejection of the null hypothesis of the Jarque–Bera goodness-of-fit test, meaning that there is no normal distribution in the data.

An innovative approach, i.e., the ‘’log t’’ regression test, has been presented by Phillips and Sul (2007) in order to evaluate the convergence methodology via a non-linear time-varying factor model and because of the need for a novel algorithm that creates clusters of groups with common or similar pathways (Du, 2017). Thus, in this research, we employed the club clustering technique proposed by Phillips and Sul (2007), to test the possible existence of club convergence and clustering. This procedure leads to a flexible panel data model, based on the decomposition of the variable under consideration into two components as follows:

Where EPit are energy poverty indicators (arrears, inadequately warm, and leaks) for the country i (i = 1,…, 31) at time t (t = 2005,…, 2022); g_it represents the systematic common components and a_it embodies transitory components, and u_t denotes a single common component and δ_it is a time varying idiosyncratic element which captures the idiosyncratic distance between the common factor u_t and the systematic part of energy poverty _it.

Phillips and Sul model the time-varying term δ_it as in a semi-parametric structure as \({\delta }_{it}={\delta }_{i}+\frac{{\sigma }_{i}{\xi }_{it}}{{L\left(t\right)t}^{\alpha }}\) where δ_i and a scale parameter σ_i are fixed across the panels, ξ_it is an i.i.d random variable with mean equal to zero and variance equal to unity across i, but weakly dependent over t, L(t) is a slowly varying function, an example of the function L(t) is log(t), which becomes infinite as t approaches infinity, and α captures the decay rate of cross-sectional variations, that is the rate of convergence of EP_it toward δ_i.

Within this framework, Phillips and Sul (2007) have coined a regression t test for the null hypothesis of convergence. The null hypothesis and its alternative (i.e., divergence) are then obtained as: \({{\rm H}_{0}: \delta }_{it}=\delta\, and\, a\ge 0\) against \({{\rm H}_{1}: \delta }_{it}\ne \delta\, and\, a<0\). For testing procedure as well as extracting information about the transition parameter δ_it Phillips and Sul (2007) used the following relative transition paths:

h_it denotes transition path associated with the behaviour of individual i with respect to the cross-section mean at time t. Supposing, that h_it converges to unity, for all i = 1, 2,…, N, with t → ∞, or if δ_it → δ as t → ∞, there is a convergence process. Therefore, the convergence hypothesis defined in Eq. 4 can be obtained as in Eq. 5, which describes that the cross-sectional variation converges to zero.

Then, the so-called ‘’log t’’ regression model can be used through the following equation:

Where \({H}_{1}/{H}_{t}\) is cross-sectional variance ration, φΤ is initial observation in the regression with φ > 0. Based on Monte Carlo experiments, the setting φ = 0.3 is suggested for a short period, i.e., when T ≤ 50 (Phillips & Sul, 2007), L(t) = log(t), and α is a parameter indicating the rate of convergence, and β is the regression parameter.

In this specific methodology, a one-sided t-test for heteroskedasticity and autocorrelation is applied to the β-coefficient. Therefore, the null hypothesis of convergence can be rejected if the t-statistic of the test does not exceed − 1.65 as a critical value at a 5% significance level. Phillips and Sul (2007) established a method that delves into finding both convergent and divergent clubs in some steps as: Step 1. Countries are sorted in descending order based on the last observation in the time series dimension of the panel, Step 2 creates the initial clusters in which countries are added until the \({t}_{\widehat{\beta }}\) for that group exceeds –1.65, Step 3 evaluates every single country that is not already in the club and to see if the new country meets the condition for membership \({(t}_{\widehat{\beta }}>-1.65)\), Step 4 refers to the “stopping rule” and the formulation of the final clubs.

This research will first present the results of a club convergence analysis on energy poverty across 31 European countries by identifying clusters of countries with similar trajectories in energy poverty reduction and alleviation. Next in order, this study will offer geographic perspectives to contextualize the regional variations in energy poverty. Finally, a robustness check using an ordered logit model based on predicted probabilities will be conducted to validate the consistency and reliability of the empirical findings.

To conclude, the present paper monitors spatial dimensions and statistically significant outcomes on energy poverty indicators, considering the development of appropriate policies to (i) the extent of energy poverty, (ii) the speed of convergence, and (iii) the level of performance in achieving SDGs. Regarding the first research question, the 31 European countries diverge on all three indicators of energy poverty.

More specifically, there is absolute level convergence in 2 of the 8 final clubs (Club 1 of Arrears and Club 4 of Inadequately Warm), and there is a conditional convergence in 5 of the 8 final clubs (Club 2, 3 of Arrears; Club 1, 2, 3 of Inadequately Warm) and in the 4 initial convergence clubs of the Leaks. Lastly, in Club 4 of the Arrears indicator, the countries show the weakest convergence. The results were further explained by implementing a robustness analysis based on ordered-logit regression results on club formations based on changes of GDPpc, a relevant and pivotal indicator for energy poverty. Regarding the second research question, the results showed that there is a core-peripheral geographical distribution on the “Arrears” indicator, whereas the indicators “Inadequately Warm” and “Leaks” have a North–South convergence pattern.

Some policy implications that could alleviate energy poverty in Europe can be linked to governmental actions, energy efficiency innovations, and societal change. Firstly, governments should incentivise renovations in poor households (i.e., to deal with leaks, damps, or rot) in tandem with the mitigation of the UHI phenomenon especially in the Mediterranean (i.e., the Southern EU), this is apparent in all of our results regarding the convergence of the energy-poor countries. Secondly, the adoption of greener renewable energy sources (e.g., creation of micro-grids) might build resilience against energy prices shocks for both North–South applications, as shown mainly in the Arrears indicators. Thirdly, the social exclusion of energy poverty (i.e., energy aporophobia) ought to be confronted with societal cooperation and governmental interventions, as social exclusion is intertwined with adverse psychological and mental effects (e.g., social isolation), these actions can be linked to the present analysis that showed common trajectories, especially in Eastern and Southern Europe. Thus, it is apparent that Agenda 2030 might be at risk on the matter of poverty eradication in all its forms (i.e., SDG 1) and that the access to clean forms of energy (i.e., SDG 7) might not have been covered wholly in Europe.

Additionally, a limitation of the present study might be the lack of data availability, except in the European territory. The three indicators have been widely collected in Europe; however, only Eurostat has gathered such data on a continent’s scale, and this methodology could not be applied to other continents due to the lack of a common energy poverty-focused dataset. This limitation, however, can stimulate future research on energy poverty in other countries or regions and this can be done by monitoring energy poverty indicators that also focus on societal-driven indicators.