The Convergence of Energy Poverty across Countries in the European Union

Abstract

With growing attention being paid to the problems of sustainable development and just transition, energy poverty emerges as a focal issue to be addressed by the Common Policy. However, the dynamics of this phenomenon across the EU still appear to be insufficiently recognised. Therefore, this study deals with the problem of energy poverty convergence and aims to test it for the 27 EU countries over the period 2010–2022. Contrary to limited studies of energy poverty convergence that use a simple indicator, it uses aggregate measures based on consensual consequential indicators of energy poverty to verify the hypothesis of convergence. Absolute and conditional beta convergence are considered. Potential determinants of energy poverty are incorporated into a model of conditional convergence. The analysis confirms the existence of beta convergence of energy poverty in the EU, indicating the progressing socio-economic cohesion of the member states. The results, thus, deliver some arguments supporting an integrative approach to the energy policy of the EU. The research reveals that, among the factors influencing EP dynamics, an important role may be attributed to technological catch-up and income distribution across a society. Sustainable development should thus be supported with energy modernisation efforts of an inclusive character.

1. Introduction

Energy poverty (EP) is a multifaceted phenomenon, with complex conceptual stems and still-evolving approaches to its definition and measurement, together with its empirically observed evolution and newly raised political challenges. The basic understanding of EP as inadequate levels of energy services [1,2,3] concerns problems of a lack of access [4], mainly discussed for developing countries, and affordability, being a focus of research for developed economies [5,6]. However, simultaneously with an emerging conciliatory perspective of energy justice, which combines energy poverty and fuel poverty issues [5], and apart from energy access and affordability, some new aspects of energy—flexibility, efficiency, needs and practices—are gaining attention [7,8]. A modern concept of energy poverty, as proposed by the United Nations [9], not only defines energy poverty as the inability to acquire a necessary amount of energy in terms of its accessibility and affordability but also underlines its reliability and high-quality environmental friendliness and safety and modernity [2,10]. Referring to Sen’s approach, it is underlined that energy poverty reflects the inability to realise essential capabilities [5,8,11] and is not limited to basic, everyday needs (such as heating/cooling, cooking, or lighting) but also covers higher, socially and culturally determined needs (e.g., education, communication, mobility) [12]. The definition of energy poverty provided by the European Commission [13] refers to access to basic energy services, which underpin a decent standard of living, health and social inclusion [14]. These different attitudes to problems of EP make it necessary to not only understand it in a material, objective context but consider its subjective aspects as well. This requires consideration of both determinants, inducing the problem of EP, and its consequences, indicating the results of the inadequate availability of energy services.

EP, as a serious socio-economic strategic challenge, has been addressed by numerous institutions aiming to reduce and mitigate it. At a global level, the United Nations has proposed its 2030 Agenda with the Sustainable Development Goals (SDGs), of which Goal 7 is focused on ‘access to affordable, reliable, sustainable and modern energy for all’ [2,5,6,15,16,17,18,19]. At the level of the European Union (EU), the issue is included in such policy initiatives as the Clean Energy for All Europeans package [20], the European Green Deal [21] with the Just Transition Mechanism and the Fit for 55 package, and it is supported by the EU’s long-term budget and the NextGenerationEU financing instrument [3,19]. Also, member states submit their National Energy and Climate Plans (NECPs) to alleviate energy poverty [8,22].

All EU countries face similar dilemmas when designing developmental policies aimed at meeting economic, social and environmental goals; however, each economy experiences some unique conditions of implementation and a different range and depth of EP problems. Moreover, the situation is evolving with different dynamics, or even different directions, in each country. This raises questions as to whether the policy measures implemented in each EU country aimed at mitigating EP should be unified and whether the potential unification should be addressed only by the strategic layout of the measures or also through the detailed instruments implemented. The answer mainly depends on the evolutionary dynamics of EP across the economies. The convergence of the phenomenon justifies searching for common solutions, or at least determining strategic directions, while the progressing divergence makes it necessary to design individual instruments addressing some unique challenges in each country.

Meanwhile, investigations into whether the EP converges across the EU countries remain a scarcity in the literature. There are some studies focused on convergence in energy consumption [4,23,24,25], environment efficiency [19,26], energy intensity [19,27] or energy transition [17], to name but a few. However, studies into energy poverty convergence—the phenomenon that merges the economic, social and environmental dimensions of sustainable development—are very rare and, to the best of our knowledge, include research by Salman et al. [28], concerning a wide range of countries across the world; González and Ibáñez-Martínand [29], referring to Latin America and the Caribbean; and Huang et al. [4] and Anastasiou and Zaroutieri [2] for the EU countries. Moreover, the results are not conclusive, which makes the diagnosis of the existence of EP convergence across the EU countries, and its political consequences, still a research gap to be fulfilled.

Therefore, the identification of evolutionary dynamics of energy poverty across the EU countries in terms of the occurrence of the convergence of this phenomenon constitutes the aim of this study. The absolute convergence of EP measured as a compound phenomenon, expressed in terms of difficulties in meeting energy needs as reported by households, is diagnosed, and its potential determining factors in terms of conditional convergence emergence are evaluated.

Theoretical complexities bring the challenge of conceptually ordering an analysis of EP dynamics, which is the value added of this study. It identifies evolutionary dynamics of EP, while putting its multidimensional causative and consequential factors in a logical order. This paper conceptualises the measurement of EP, adopting a comparative approach, which is reflected by an analysis of EP dynamics in terms of convergence across the EU countries. It uses an attitude that allows its causes and effects to be considered respectively—objectively as socio-economic determinants at a country level and subjectively as declared by households. Moreover, this study bases its analysis on an aggregate measures to better reflect the EP phenomenon while testing the convergence hypothesis, deliberately distinguishing the EP determinants in the model of conditional convergence. Although causative and consequential indicators are of high importance for understanding both the capacity of a state to tackle the issue and the actual situation of the issue in the state [8], the unifying ordered approach used in this study is not common in the literature.

Research into the convergence of the phenomenon across a group of countries is rare, although it may finally deliver valuable arguments to properly form a common energy policy. This study deals with the diagnosed literature gap and tests EP convergence across the EU countries to give a scientific basis for policy. It allows us to better understand the dynamics of the EP in conditions of progressing socio-economic integration. It is assumed that identifying EP convergence supports an integrative approach to energy policy, while diagnosing EP divergence can allow for arguments for a unique policy conducted by each member state to be delivered. The verification of the hypothesis about the convergence of EP is a precondition for rational policy design within the EU. Hence, this study is of high importance not only from a diagnostic point of view but also for energy policy formulation.

2. Literature Review and Conceptual Framework of the Research

Traditionally the phenomenon of convergence is referred to in economic terms and understood as a process of limiting the gap in a level of GDP per capita between economies [2,30,31,32,33,34,35,36]. When convergence occurs, poor economies experience higher economic growth than rich ones, the differences in GDP per capita are reduced and the economies tend to achieve steady-state equilibrium. This process is assessed through concepts of beta or sigma convergence. The former tests the hypothesis about a significant negative relationship between the economic growth rate and the initial level of GDP per capita, therefore verifying whether poor countries develop faster than rich ones. It is considered either in absolute or conditional form (when controlling for some specific characteristics of the economies and investigating factors potentially influencing the growth). The latter concept tests whether the variation in GDP per capita between economies reduces and captures the cross-sectional equilibrium over time. The convergence is sometimes analysed in terms of stochastic convergence, identified when the GDP per capita of an economy relative to the benchmark economy appears as a stationary process and when long-term economic growth does not depend on country-specific factors. If the convergence is not identified for the entire sample of countries, it is also possible to search for convergence clubs [2,36,37,38].

Recently, the traditional neoclassical economic theory of convergence has been increasingly used to investigate the evolution of socio-economic phenomena other than the income level. Research is conducted concerning not only the economic but also social and environmental dimensions of development processes. There are some studies covering problems of environmental convergence referring to the worldwide convergence of energy intensity levels [27], convergence of sustainable energy transitions of 82 economies of selected income levels [17] or European convergence of environmental technical efficiency, which partially refers to energy poverty [19,26]. These studies find convergence among EU countries (in terms of sigma, as well as absolute and conditional beta convergence) [19,26] and conditional worldwide convergence [27]. Moreover, there are also studies that deal with social problems and investigate income inequality convergence, usually confirming the hypothesis in the countries across world [39,40], within developed countries [36] or in the EU [41,42,43].

Nevertheless, there are only a few papers that strictly deal with the dynamic patterns of EP. In this vein, energy poverty convergence is understood—parallel to traditional income convergence—as a process in which EP declines faster in countries with a high initial level, and the gap in EP among different countries is narrowing [4]. However, the studies do not provide consistent results. Huang et al. [4] supported the existence of both conditional and unconditional β-convergence of energy poverty across 28 EU countries, while Anastasiou and Zaroutieri [2] rejected it for 27 EU member states, identifying convergence clubs instead. On a broader scale, Salman et al. [28] found divergence in energy poverty across 146 countries and the formation of six convergence clubs. Convergence was also identified for Latin America and the Caribbean [29]. The studies differ not only in how they test the convergence hypothesis but also in their approach to EP measurement. They either use the multidimensional energy poverty index [28], alternative single subjective measures of energy poverty [2,4] or some single measures referring to energy access, quality and affordability [29], which represent a broader but less consistent scope of EP indicators than in this study. Therefore, it seems necessary to use a conceptually ordered approach to EP (distinguishing its causes and manifestations) when testing the existence of convergence.

EP measurement is a challenging task; even though numerous indicators and approaches have been proposed in the literature, there are still serious limitations in the availability of empirical data. Moreover, different measures are perceived as expressing specified aspects of EP, and they often concern different population segments and diverge in terms of incidence, seasonality, cross-country comparability and persistence over time [3,44]. Generally, three methods have been identified for measuring EP—direct measurement, an expenditure approach and consensual (subjective) approach—all of them having distinct advantages and drawbacks [1,4,8,18,22]. The diversity of approaches and indices used to measure EP is well expressed in the review study by Siksnelyte-Butkiene [10], who groups the approaches into three subcategories focusing on energy access, energy poverty and energy vulnerability, with distinct attitudes concerning developed and developing countries, and identifies such categories of indices as income indicators, expenditure indicators, energy price indicators, energy consumption indicators, household characteristic indicators, dwelling characteristic indicators, comfort indicators and access to the energy indicators that were assigned to three sustainability dimensions (economic, social and environmental) [10]. In another study, Igawa and Managi [6] categorise EP indicators into five types, e.g., connection-based, energy consumption-based, energy service-based, energy expenditure-based and consensual-based, while Menyhert [3] pays attention to their various characteristics—monetary and non-monetary, subjective and objective, qualitative and quantitative, primary and secondary, absolute and relative kind—concluding that many forms of EP still remain hidden under the existing measurement framework.

Additionally, apart from single indicators, many researchers have developed some composite multidimensional energy poverty indices, taking into account that no single method can capture the multifaceted nature of EP. Among these composite indices, there are, e.g., the Multidimensional Energy Poverty Index (MEPI) by Nussbaumer et al. [45], the Energy Poverty Multidimensional Index (EPMI) by Bollino and Botti [46], the Composite Energy Poverty Index (CEPI) by Bouzarovski and Herrero [47] and the Multidimensional Energy Poverty Index (MEPI) by Sokolowski et al. [8,10,14,22,48]. However, this multidimensional approach carries the risk of confusing indices specifying determinants or vulnerability factors of energy poverty that reflect causes of the problem with indices of EP consequences, that reflect a manifestation of the phenomenon and the difficulties suffered by those identified as energy-poor.

Meanwhile, the literature distinguishes numerous determinants of EP as well, identified both at country and individual levels. The most-often analysed consensual factors of EP cover income (at a household or a country level), infrastructural conditions influencing the efficiency of energy usage (of buildings and appliances or performance of an economy as a whole, reflected by energy productivity or intensity) and energy prices [1,8,14,18,22,47]. There are also studies considering climate conditions [2,6], income inequality or poverty [4,6] and urbanisation [4,6,47], as well as a set of socio-economic features of households determining individual energy needs [1,6,18,47]. It is claimed that income at a household level can either directly (ability to pay fuel bills) or indirectly (ability to invest in energy-efficiency improvements) influence EP [6,18], and, at a country level, it differently influences each dimension of EP in terms of accessibility, reliability and affordability [6], with high-income economies still having lower EP [49]. Generally, a higher income is expected to decrease EP, while how the income is distributed in a society is also important, as high-income inequality can aggravate subjective affordability for all households [6]. Our conceptual framework considers the main determinants of EP when analysing the conditional convergence of EP expressed by its subjectively perceived effects, taking advantage of some previous research [8] that concludes that the capacities of EU countries to address the EP issue often result in different consequential aspects of EP. Therefore, the causes and effects of EP cannot be perceived as substitutes for EP evaluation, and their conceptualisation is an important precondition for any analysis.

This study focuses on EP convergence evaluated in the form of the beta convergence hypothesis, which assumes that the precondition for achieving similarity between economies is that the countries with initially higher levels of EP are more successful in alleviating the problem. Such situations could lead to growing cohesion in sustainable development and is recognised as favourable for European integration.

When testing beta convergence in absolute terms, we use consensual measures that reflect problems experienced by people not able to satisfy their energy needs and, thus, are specified as results or consequences of EP. In order to capture the compound character of the problems, we use a multidimensional measure of EP (Synthetic Energy Poverty Measure (SEPM) and Compound Energy Poverty Measure (CEPM)) that aggregates three commonly used single indicators of EP, and, additionally, we also test the convergence hypothesis using single measures of the consequential type. In the next step, we also test conditional convergence by controlling for the most commonly used determinants of EP that are perceived as causes of EP. Therefore, we distinguish between two kinds of EP indicators—determinants and consequences—and incorporate them in our convergence analysis either as directly reflecting the EP phenomenon or as causal factors conditioning the convergence process (Figure 1).

3. Materials and Methods

This study focuses on the EU as the most advanced integration example in the world. It is expected that, along with progressing economic, social and political integration across the member states, EP convergence will appear as well, and the countries will increase their multidimensional cohesion.

The research focuses on EP convergence and covers three steps of analysis:

• Diagnosis of EP across EU countries using aggregate measures of consequential factors;
• Verification of EP absolute beta convergence based on the aggregate measure calculated in the first step;
• Verification of EP beta convergence in terms of its conditional form considering the main factors potentially determining EP.

In the first step, the EP and its changes are evaluated in terms of consensual-based subjective indicators reflecting the self-reported problem across the population of the 27 countries of the European Union in the period 2010–2022. We used a dataset from Eurostat EU-SILC (European Union Statistics on Income and Living Conditions) and derived three commonly used indicators, listed as follows:

• Housing cost overburden rate [50]—which is the percentage of the population living in households where the total housing costs represent more than 40% of disposable income [51] (henceforth: cost_over);
• Inability to keep home adequately warm [52]—which expresses the percentage of the population self-reporting the problem (henceforth: in_warm);
• Arrears on utility bills [53]—which reflects the percentage of the population unable to pay utility bills for their main dwelling on time due to financial distress (henceforth: arr_ub).

The higher the value of any of the three variables, the higher the EP level.

It is worth stressing that the three consequential indicators are collected by Eurostat within nationally conducted surveys based on common concepts and definitions. It allows for catching subjectively reported attitudes about EP. The self-assessment approach has its disadvantages with a possible error of exclusion, e.g., when households do not perceive themselves as energy-poor even though they are or do not want to report their difficult situation. The answers thus may be influenced by culturally specific perceptions. On the other hand, the survey allows for capturing immaterial social factors conditioning EP, which are omitted with ‘hard’ data based, e.g., on the direct measurement of thermal conditions. EP as a social phenomenon may be perceived as conditioned by the time and place of living, and these circumstances are reflected by the considered data.

However, there are some specific issues that one should be aware of when considering each detailed indicator as reflecting EP. Some bias can be connected with the broad definition of housing costs, which covers paid monthly expenses associated with the right to live in a dwelling, not only energy costs [51]. A similar limitation refers to utility bills, which include bills for heating, electricity, gas, water, sewage, rubbish, etc. [54]. The definitions thus do not allow for a focus on the difficulties with addressing strictly energy needs but rather reflect more general problems of poverty. Nevertheless, problems with material maintenance are revealed as a main correlate of EP. Moreover, it is assumed in the questionnaire that the household is in the same situation independent of the source of paying the bills (through borrowing or its own resources) [54]. This methodological solution blurs the perception of poverty tensions and limits its identification to the actual situation without capturing its cumulative nature. Finally, the respondents are asked about their abilities to pay, irrespective of their actual energy needs [54], which pay attention to financial aspects of EP rather than its broader picture. These limitations make the indicators only roughly reflect the phenomenon of EP, but they remain the main source of enriching our understanding of the problem across time and countries in a comparable manner. Even though numerous attempts have been taken, there is still no other available comprehensive longitudinal cross-sectional database gathering data on EP.

In this study, these three indicators will later be used to construct a synthetic measurement of EP, which is based on two alternative methods—the zero-unitarisation [55] and Hellwig methods [56]—both used for dynamic comparisons.

In the zero-unitarisation method, the three variables are normalised using the following formula:

$${\mathrm{z}}_{\mathrm{n}\mathrm{i}\mathrm{t}}={\displaystyle \frac{{\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}-\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}\right)}{\max \left({\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}\right)-\min \left({\mathrm{x}}_{\mathrm{i}\mathrm{t}}\right)}}$$

(1)

where x_nit is the value of the n-th variable (n = 1,2,3) in country i (i = 1…,27) in time period t (t = 1,…13, period: 2010–2022). The formula uses the difference between the actual value of the variable and the lowest possible value in the group of countries in the researched period, expressing the distance to the most desired situation. The higher the distance, the more severe the EP problem is.

Then, the synthetic measure, SEPM, is calculated using the following formula:

$${\mathrm{S}\mathrm{E}\mathrm{P}\mathrm{M}}_{\mathrm{i}\mathrm{t}}={\displaystyle \frac{\sum _{\mathrm{n}=1}^3{\mathrm{z}}_{\mathrm{n}\mathrm{i}\mathrm{t}}}{3}}$$

(2)

The formula aggregates the individual indicators of EP. The SEPM can take values from the range [0;1]. The closer the value to 1, the higher the level of EP is.

In the Hellwig method, normalisation is carried out with the following formula:

$${\mathrm{z}}_{\mathrm{n}\mathrm{i}\mathrm{t}}={\displaystyle \frac{{\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}-\overline{{\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}}}{{\mathrm{S}(\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}})}}$$

(3)

where $\overline{{\mathrm{x}}_{\mathrm{n}\mathrm{i}\mathrm{t}}}$ is the arithmetic mean of the variable, and S(x_nit) is the standard deviation of the variable.

Then, the Euclidean distance d_it to the maximum z_nit (as the maximal values reflect the highest EP) is calculated for each country i in each year t.

Finally, the compound energy poverty measure, CEPM, is calculated by the following formula:

$${\mathrm{C}\mathrm{E}\mathrm{P}\mathrm{M}}_{\mathrm{i}\mathrm{t}}=1-{\displaystyle \frac{{\mathrm{d}}_{\mathrm{i}\mathrm{t}}}{\mathrm{d}}}$$

(4)

where d = $\overline{{\mathrm{d}}_{\mathrm{i}\mathrm{t}}}$ + 2S(d_it), $\overline{{\mathrm{d}}_{\mathrm{i}\mathrm{t}}}$—the arithmetic mean of the Euclidean distance and S(d_it)—the standard deviation of the Euclidean distance.

The CEPM usually takes values in the range [0;1]. The higher the CEPM is, the higher the level of EP.

The two alternative methods of aggregation deliver two measures of EP that are used to test the convergence hypothesis, considering the robustness of the results.

In the second step of this study, the beta convergence hypothesis is verified, assuming that countries with a higher initial level of EP tend to decrease it at a faster pace. The initial verification of absolute convergence requires the estimation of this Equation:

$${(\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}}-{\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}-1})={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(5)

where EP_i,t is the EP index of country i in period t (measured alternatively by SEPM or CEPM), ε_it is the residual term, β₀ is the intercept and β₁ is the parameter of interest. β₁ identifies convergence when it exhibits a negative and significant value—formally, the convergence hypothesis is not rejected if β₁ < 0.

• In the third step in this paper, we examine the conditional beta convergence of EP to identify potential factors influencing EP. To move into a conditional test of convergence, some variables specifying the potential determinants of EP are added to Equation (5). Based on the literature review, we considered variables referring to income level, income distribution, energy prices, and efficiency of energy usage. Our models consider data derived from the Eurostat database and cover the period 2010–2022, e.g., real GDP per capita [57]—at market prices in euro per capita as chain-linked volumes (2010)—which reflects the level of a country’s development and the average income of its population, expressing material living standards (henceforth: gdp_pc);
• The Gini coefficient of equivalised disposable income [58]—this indicates the distribution of income across the population and how severe income inequality is on a scale [0;100], where 0 means equal distribution, and the higher the value, the higher the income inequality (henceforth: gini);
• Electricity prices per kilowatt-hour (all taxes and levies included) in PPS for household consumers (band DC) in the second half of a year [59]—this indicates the costs of access to energy for typical consumers (henceforth: e_price);
• Energy intensity of GDP in PPS in KGOE per thousand euros in PPS [60]—this indicates how efficiently the economy works in terms of its energy performance, and it reflects the state of the infrastructure, which includes a country’s technological advancement and also the technical conditions of buildings (henceforth: en_int).

Finally, the model for conditional convergence takes the following form:

$${(\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}}-{\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}-1})={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\beta }}_2{\mathrm{g}\mathrm{d}\mathrm{p}\_\mathrm{p}\mathrm{c}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_3{\mathrm{g}\mathrm{i}\mathrm{n}\mathrm{i}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_4{\mathrm{e}\_\mathrm{p}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{e}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_5{\mathrm{e}\mathrm{n}\_\mathrm{i}\mathrm{n}\mathrm{t}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(6)

where β₂–β₅ are estimated coefficients for the newly added variables, and the rest of the notations are as above.

The coefficients of all the models are estimated for robustness checking alternatively by OLS, FE, RE, 1-step GMM and 2-step GMM methods. The logarithmic forms of the variables are used to address the problems with their distribution.

4. Results

A starting point in our assessment of EP across the EU countries (Figure 2,

Table 1. EP across the 27 EU countries measured by SEPM and CEPM.

	SEPM		CEPM
	Avg. 2010–2022	Change 2022–2010	Avg. 2010–2022	Change 2022–2010
Belgium	0.1133	−0.0327	0.1699	−0.0252
Bulgaria	0.5097	−0.2572	0.4752	−0.0682
Czechia	0.0911	−0.0513	0.1525	−0.0471
Denmark	0.1460	−0.0355	0.2031	−0.0244
Germany	0.1371	−0.0065	0.1975	−0.0071
Estonia	0.1014	−0.0604	0.1413	−0.0359
Ireland	0.1478	−0.0217	0.1704	−0.0147
Greece	0.6208	0.2056	0.5638	0.1323
Spain	0.1681	0.0586	0.2171	0.0567
France	0.1031	0.0358	0.1511	0.0400
Croatia	0.2552	−0.1936	0.2314	−0.1399
Italy	0.1830	−0.0728	0.2297	−0.0531
Cyprus	0.2460	−0.1121	0.2548	−0.0767
Latvia	0.2238	−0.2286	0.2456	−0.1721
Lithuania	0.2446	−0.1361	0.2841	−0.1305
Luxembourg	0.0660	0.1056	0.1243	0.1014
Hungary	0.2306	−0.1649	0.2429	−0.1145
Malta	0.1294	−0.0496	0.1653	−0.0476
Netherlands	0.0988	−0.0198	0.1620	−0.0151
Austria	0.0722	−0.0209	0.1297	−0.0158
Poland	0.1549	−0.1527	0.1951	−0.1275
Portugal	0.1943	−0.0715	0.2459	−0.0555
Romania	0.2978	−0.1503	0.3038	−0.1246
Slovenia	0.1566	−0.1056	0.1679	−0.0570
Slovakia	0.1109	−0.0547	0.1598	−0.0457
Finland	0.0815	−0.0007	0.1217	0.0051
Sweden	0.0813	0.0028	0.1397	0.0100
Avg.	0.1839	−0.0589	0.2165	−0.0390

Source: own study based on Eurostat data.

) is an observation that the most severe energy problems are faced in Greece and Bulgaria, where aggregated measures of EP are well above the EU average (which was 0.22 for CEPM and 0.18 for SEPM). EP also appears as a serious threat in such countries as Croatia, Cyprus, Latvia, Lithuania, Hungary, Portugal and Romania, as well as Spain and Italy. This is in line with the diagnosis of Bouzarovski and Herrero [47] that the EP distribution in Europe is specified by the core–periphery pattern, with relatively higher EP in the South and Central and Eastern European (CEE) countries. Nevertheless, there are some economies such as the Czech Republic or Slovakia that break out of the pattern and align with the core group rather than with the peripheral CEE. The below-average level is also achieved by Estonia, Poland and Slovenia. Still, the states with less-serious problems of EP are Finland, Luxembourg, Austria, Sweden, France, Netherlands, Malta, Belgium, Ireland, Germany and Denmark. The observations hold true for both the SEPM and CEPM, and the measures are strongly correlated at a level of 0.987, which shows the robustness of the EP measurement to a change in the aggregation method. Both the SEPM and CEPM measures of EP are alternatively used in the following EP convergence analysis.

Moreover, for most of the period (in all the years from 2013 to 2021), the average level of EP across the EU-27 declined, showing serious improvement between 2010 and 2022 (

Table 2. EP in the period 2010–2022—the average for the UE-27 measured by SEPM and CEPM.

	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
SEPM	0.2037	0.2092	0.2259	0.2322	0.2257	0.2088	0.1931	0.1735	0.1619	0.1497	0.1352	0.1273	0.1447
CEPM	0.2278	0.2364	0.2511	0.2538	0.2481	0.2349	0.2243	0.2096	0.1992	0.1900	0.1779	0.1726	0.1888

Source: own study based on Eurostat data.

). EP appears to be a strongly cyclical phenomenon, as it increased in 2010–2013 when the economic growth across the EU was slow, while, after an upturn in 2013, it started to decline as the economic situation was improving. Although essential changes in all spheres of human existence were initialised by the COVID-19 pandemic, the measures were not affected by the pandemic, and the decreasing trend continued in 2020–2021. The last year of the period—2022—marked by the energy crises connected with the Russian military aggression against Ukraine, appeared to be specified by EP growth. Even though our analysis covers only the first year of the conflict, it essentially influenced EP in the EU countries through drastic changes in energy supply directions, market shortages of energy sources and a sharp increase in market energy prices. It appears that such perturbations particularly affect those most at risk of EP.

Comparing individual countries in the analysed period (Table 1), the situation improved the most in economies such as Bulgaria, Latvia, Croatia, Hungary, Poland and Romania, as well as in Lithuania (reflected mainly by CEPM) and Bulgaria (with the highest decrease reflected by SEPM). All of them belong to the group of CEE countries reflecting the great socio-economic progress of the ‘new’ member states, also concerning EP alleviation. On the other hand, EP increased in Greece and Spain, countries already experiencing severe EP problems, as well as in countries with relatively low EP levels, Luxembourg, Sweden, France and also Finland (only when measured by CEPM). These comparisons reveal that it is the south of Europe where the growing tensions connected with meeting the energy needs of society are mainly faced.

Our research confirmed the existence of EP absolute convergence across the EU 27 countries (

Table 3. Estimations of the EP convergence models with SEPM for the 27 EU countries in 2010–2022.

	Absolute Convergence					Conditional Convergence
	OLS	FE	RE	1-Step GMM	2-Step GMM	OLS	FE	RE	1-Step GMM	2-Step GMM
d_l_SEPM
l_SEPM_1	−0.0414 * (0.0210)	−0.1141 *** (0.0322)	−0.0414 ** (0.0210)	−0.6684 *** (0.1077)	−0.6383 *** (0.1209)	−0.0448 (0.0314)	−0.3162 *** (0.0485)	−0.0448 (0.0314)	−0.7446 *** (0.0979)	−0.7051 *** (0.1151)
l_gdp_pc						0.0504 *** (0.0175)	−0.8579 *** (0.2740)	0.0504 *** (0.0175)	−1.1238 *** (0.3764)	−0.4674 (0.4769)
l_gini						0.2030 *** (0.0703)	0.3283 (0.2236)	0.2030 *** (0.0703)	0.8787 * (0.5164)	0.8183 * (0.4241)
l_e_price						0.0906 * (0.0511)	0.2290 *** (0.0820)	0.0906 * (0.0511)	0.1051 (0.1231)	0.1468 (0.1339)
l_en_int						0.0430 (0.0265)	−0.0521 (0.1760)	0.0430 (0.0265)	−0.9906 *** (0.2411)	−0.8405 *** (0.3015)
d_l_SEPM (−1)				−0.0891 (0.1663)	−0.1415 (0.1738)				−0.1398 (0.1244)	−0.1807 (0.1368)
const	−0.1047 ** (0.0410)	−0.2415 *** (0.0606)	−0.1047 ** (0.0410)	−0.0374 *** (0.0110)	−0.0343 *** (0.0124)	−1.3651 *** (0.2924)	7.4420 ** (3.2885)	−1.3651 *** (0.2924)	−0.0577 *** (0.0115)	−0.0639 *** (0.0131)

HAC robust standard errors in parentheses; * 0.1, ** 0.05, *** 0.01 statistical significance. Source: own study based on Eurostat data.

and

Table 4. Estimations of the EP convergence models with CEPM for the 27 EU countries in 2010–2022.

	Absolute Convergence					Conditional Convergence
	OLS	FE	RE	1-Step GMM	2-Step GMM	OLS	FE	RE	1-Step GMM	2-Step GMM
d_l_CEPM
l_CEPM_1	−0.0354 * (0.0196)	−0.1256 *** (0.0298)	−0.0354 * (0.0196)	−0.6998 *** (0.0928)	−0.6529 *** (0.1108)	−0.0395 (0.0305)	−0.3448 *** (0.0576)	−0.0395 (0.0305)	−0.7588 *** (0.1075)	−0.7661 *** (0.1190)
l_gdp_pc						0.0358 *** (0.0098)	−0.5897 *** (0.1895)	0.0358 *** (0.0098)	−0.6230 ** (0.2632)	−0.3678 (0.3031)
l_gini						0.1251 *** (0.0420)	0.2195 (0.1516)	0.1251 *** (0.0420)	0.4973 * (0.2898)	0.5983 ** (0.2907)
l_e_price						0.0644 * (0.0352)	0.1587 *** (0.0537)	0.0644 * (0.0352)	0.0818 (0.0845)	0.0886 (0.0830)
l_en_int						0.0382 ** (0.0176)	−0.0355 (0.1139)	0.0382 ** (0.0176)	−0.6455 *** (0.1273)	−0.5980 *** (0.1679)
d_l_CEPM (−1)				0.0111 (0.1434)	−0.0455 (0.1528)				−0.0568 (0.1270)	−0.1431 (0.1402)
const	−0.0724 ** (0.0337)	−0.2179 *** (0.0481)	−0.0724 ** (0.0337)	−0.0250 *** (0.0065)	−0.0223 *** (0.0068)	−0.9418 *** (0.1649)	4.9931 ** (2.3165)	−0.9418 *** (0.1649)	−0.0406 *** (0.0064)	−0.0432 *** (0.0068)

HAC robust standard errors in parentheses; * 0.1, ** 0.05, *** 0.01 statistical significance. Source: own study based on Eurostat data.

). Both for models of beta convergence using SEPM and CEPM measures, the β₁ coefficient appeared to be negative and statistically significant, independent of the estimation method used. The hypothesis assuming that countries with initially higher levels of EP were tackling the problem efficiently enough to limit EP faster than the economies that did not experience such severe EP could thus be confirmed for the EU.

This research also identified conditional beta convergence as controlling for income level and distribution, energy prices and energy intensity (Table 3 and Table 4). In the cases of conditional convergence, the β₁ coefficient maintained its negative sign, although it lost its significance in some models (however, when comparing OLS vs. FE vs. RE models, the F test, Breusch–Pagan and Hausman tests indicated the choice of FE models, where the coefficient was still significant, and Sargan tests indicated the appropriateness of the 2-step GMM models, where β₁ significance was maintained). The conclusion reflects growing cohesion and adjustments to common SDGs across EU countries.

When the 2-step GMM model, which allows for addressing some problems of endogeneity and autocorrelation, is analysed (for SEPM and CEPM), it not only confirms the convergence of EP but it also points at the energy intensity of an economy as an important factor influencing the dynamics of EP alleviation. It suggests that the technological catch-up, which allows for the improvement of energy performance, may be crucial in limiting EP. Investments in modern technical infrastructure in industry, transportation and energetic transformation in social and private infrastructure could be favourable not only for the economy but for the society and environment as well. These models also confirm that income inequality interacts with EP, as a multidimensional poverty phenomenon covers its energy, as well as income aspects. In countries with higher income inequality, the poverty rate is usually higher, and, thus, it is also more challenging to deal with insufficient energy availability.

As the aggregated measures of EP give a balanced view of the problem and some researchers have proven that each EP indicator can reflect different aspects of the problem [3], we also tested absolute beta convergence concerning the simple indicators used to construct our aggregate measures of SEPM and CEPM (

Table 5. Estimations of the EP absolute convergence models with simple indicators for the 27 EU countries in 2010–2022.

	OLS	FE	RE	1-Step GMM	2-Step GMM
	d_l_cost_over
l_cost_over_1	−0.0550 ** (0.0256)	−0.3299 *** (0.0494)	−0.0550 ** (0.0256)	−1.0569 *** (0.0954)	−1.0426 *** (0.1108)
d_l_cost_over (−1)				−0.1627 * (0.0934)	−0.1672 * (0.0994)
const	0.0971 * (0.0504)	0.6563 *** (0.1006)	0.0971 * (0.0504)	−0.0384 *** (0.0121)	−0.0347 *** (0.0129)
	d_l_in_warm
l_in_warm_1	−0.0589 *** (0.0124)	−0.2819 *** (0.0532)	−0.0589 *** (0.0124)	−0.9039 *** (0.1586)	−0.9007 *** (0.1723)
d_l_in_warm (−1)				−0.0802 (0.1301)	−0.0815 (0.1390)
const	0.1055 *** (0.0254)	0.5273 *** (0.1007)	0.1055 *** (0.0254)	−0.0359 ** (0.0156)	−0.0356 ** (0.0180)
	d_l_arr_ub
l_arr_ub_1	−0.0350 ** (0.0150)	−0.1796 *** (0.0544)	−0.0350 ** (0.0150)	−0.9812 *** (0.0950)	−0.8905 *** (0.1243)
d_l_arr_ub (−1)				−0.0241 (0.1129)	−0.0733 (0.1360)
const	0.0371 (0.0306)	0.3343 *** (0.1118)	0.0371 (0.0306)	−0.0534 *** (0.0105)	−0.0489 *** (0.0118)

HAC robust standard errors in parentheses; * 0.1, ** 0.05, *** 0.01 statistical significance. Source: own study based on Eurostat data.

). Once again, the hypothesis was confirmed in every case—concerning the three simple indicators, as well as different methods of model estimation. The β₁ coefficient was significantly negative, proving that the EU countries that experience a higher EP—as reflected in the share of the population that suffers housing cost overburden, is not able to keep the home adequately warm or is experiencing arrears on utility bills—also reduce it faster. Therefore, it could be summarised that EP in its consequential aspects is converging across the 27 EU countries.

5. Discussion

The presented results are in line with the literature identifying the energy poverty divide not as a problem of the ‘old’ and ‘new’ member states only but rather as a core–periphery pattern also including some ‘old’ EU economies as laggards [47]. Many researchers identified that some of the ‘old’ EU countries, especially in the south of Europe, have a more severe EP problem than some of the ‘new’ member states [14,61]. Generally, it is found that the EP level is lower in the northern and western countries, while it concentrates in the southern and eastern countries [8], and the spatial context distinguishes Northern, Central and Mediterranean Europe [62]. A similar description could also be derived from our aggregate measures.

Moreover, studies on EP dynamics show that EP is limited mainly in the ‘new’ member states, while its growth is observed in the ‘old’ members [14]. The patterns are also, to some extent, supported with our analysis of the SEPM and CEPM dynamics, finding that EP problems were increasingly concentrated in the European South, while they were limited with the highest speed in the CEE countries.

Finally, the scarce research into EP convergence across the EU indicates opposing results—either that the convergence appears and is robust, as shown by a different convergence analysis method [4] or that it concerns only some convergence clubs [2]. The findings of our study support the identification of convergence by Huang et al. [4] rather than limiting it to specified clubs, as in [2]. The conclusion delivers an important argument justifying integrating efforts in terms of EP policy. It supports the idea of unifying energy goals and attitudes across the EU countries. It suggests a need to exchange experiences and share ‘good practices’ to alleviate EP and to develop a common energy policy.

It is also important to stress some methodological differences between our study and other research on EP convergence. While we adopted the method of beta analysis convergence to verify its existence across EU countries, concerning some aggregate measures (SEPM or CEPM), Anastasiou and Zaroutieri [2] referred to individual indicators, and, after rejecting the convergence hypothesis, they adopted the Phillips and Sul data-driven algorithm to identify convergence clubs. On the other hand, Huang et al. [4] identified convergence using panel quantile regression, which allowed them to reach conclusions about a heterogeneous rate of convergence, with countries with high EP levels reducing it more quickly. Nevertheless, they also based their analysis on individual indicators of EP. Although our study differs from these analyses considering EP measurement or estimation techniques, it is also possible to point to some common findings of each. The most important is to identify EP convergence at least in some groups of EU countries.

Our study differs from previous convergence research for the EU, as it uses the aggregated measure to test the convergence hypothesis. This approach allows us to capture the dynamics of the compound phenomenon of EP in a more complete way. We focus on beta convergence in its absolute and conditional forms, as it is a precondition for finding any long-term diminishing gap between economies. Still, we deliberately distinguish between the consequential factors that reflect the EP phenomenon and the determinants of the EP, which we incorporate into our model of conditional convergence. The findings of the study not only confirm the convergence of EP (in terms of its results) between the EU countries but also draw attention to such factors of EP as energy intensity and income inequality, which are sometimes omitted or substituted by more obvious determinants of EP, such as income level and energy prices. Improving social cohesion and investing in modern technologies thus appear as mechanisms allowing for EP alleviation. The findings suggest that it is not enough to increase the average income of society and to publicly control the energy prices for consumers to reduce the extent of EP. More efficient mechanisms of EP policy can be found in stimulating energy modernisation efforts undertaken by individuals, as well as local government units and businesses. Instruments of EP policy should simultaneously be cohesive with actions aimed at limiting social inequality. More equal income distribution enhances the spillover effects of energy-saving solutions, which could be favourable for the economy, society and environment.

Nevertheless, all of the policy actions face an ‘energy dilemma’ or even an ‘energy trilemma’ when they are, at the same time, aimed at meeting economic, social and environmental needs. The policy has to balance the energy affordability issue and actions to address climate change [6], and, as the Council of the European Union declared, the energy transition has to be ‘fair and inclusive’ [19]. The goals of energy security (meeting the energy demand of the whole economy), energy equity (ensuring that the energy supply is affordable for the entire population) and environmental sustainability (meeting the resource and service needs of current and future generations without losses to ecosystems) have to be simultaneously achieved [19], which appears to be a challenging task and requires carefully designed measures. Meanwhile, when climate change actions are implemented through higher energy taxes, carbon pricing and fossil fuel subsidy phase-outs, a disproportionately heavier burden is placed on households with low income, and unwanted redistributive consequences are likely to result. Thus, the trade-offs between EP mitigation and decarbonisation efforts make the policy challenging [6,18]. However, some research shows that a rational policy that allows for meeting the net-zero target, while also reducing inequality, is possible. It is suggested that as decarbonisation actions can increase inequality, with low-income households being more seriously affected, it may be balanced by using carbon revenues as lump-sum transfers to households without hampering goals for equity and development [22]. Moreover, the literature delivers some arguments that financial inclusion initiatives may support sustainable development through increase in energy efficiency [63]. There are also some studies stressing the role of public support in the process of energy transition with a mean of renewable energy development, suggesting that such a policy allows for simultaneously meeting energy demands and mitigating environmental impact [64]. The findings of our study support the idea of limiting income inequality as an important tool in alleviating EP. Moreover, more equal income distribution appears to induce dynamics of EP reduction while it co-occurs with energy modernisation based on new energy-saving and environmentally friendly technologies.

6. Conclusions

The main finding of this study is that EP converges across EU countries; that is, states with initially higher levels of EP are able to alleviate the problem more dynamically. This observation provides a basis for positively assessing the efforts taken by the more disadvantaged countries concerning their struggle with EP and allows for the chance of enhancing socio-economic and environmental cohesion between the member states.

Nevertheless, there appears to be a serious threat of limiting the dynamics of positive change after the exhaustion of financial support from the EU’s common sources and a potential failure to activate the endogenic mechanisms supporting sustainable development. Also, the experiences of the countries more advanced in EP alleviating suggest that limiting EP beyond some level could appear impossible. When the extent of EP concerns only a limited share of the population, the social pressure on public policy diminishes, and a common perception of the problem may turn into a negative and exclusive narrative. When mechanisms of protection against the EP risk are not incorporated into objective long-term instruments of social policy, EP could transform into a vicious cycle of inherited marginalisation. Thus, the challenge for the EU countries is not only to set ambitious targets of EP alleviation but also to identify the best solutions of universal character that could be flexibly adopted in each of the member states.

The findings of this study also confirm the role of technological catching-up in limiting gaps between EU countries in terms of EP. It appeared that less technologically advanced economies were able to improve their EP situation more dynamically. The energy modernisation of infrastructure in all its forms—private houses and equipment, as well as public infrastructure used to deliver social services, more advanced solutions in transportation and technical infrastructure used in industry—can constitute a serious trigger to reduce energy input in order to obtain more social gains in terms of production, income and environmental protection and thus also limit EP. Moreover, this research indicates that limiting EP cannot omit its interrelations with income inequality, and social support policies should address the problems of poverty in its multidimensionality.

This presented study is nevertheless limited in its diagnosis of the EP problem and convergence testing, both by data limitations and by the preliminary character of the convergence analytical approach. First, EP as a compound phenomenon needs to be considered in more aspects that reflect the problems of people struggling to meet their energy needs. This study used aggregate measure of only three consequential EP indicators, which could be browsed in future research. Nevertheless, the available data describing the phenomenon are limited to the consensual approach, which is extremely important in expressing the social character of the phenomenon but should be supplemented by other objective indicators of EP based on the expenditure and income approach, as well as its direct measurement to give a more comprehensive picture of the problem. With the development of common initiatives aimed at creating such a database, the understanding of this problem will seriously improve. Second, the beta convergence hypothesis may be treated as a preliminary step of investigating the process of EP dynamics. Future research could develop the idea by sigma and stochastic convergence analysis, as well as testing the influence of other factors potentially determining the EP level in a country. Moreover, an interesting space for future analysis is the differences in the speed of EP convergence across countries and the determinants of such differences. Exploring, in more detail, the potential factors decisive for the dynamics of EP alleviation is of high importance for energy policy formulation.