The Role of Poverty and Gender in Shaping Households’ Energy Consumption Patterns in Selected European Countries

Abstract

In the context of Sustainable Development Goals, declining poverty (Goal 1), achieving gender equality (Goal 5), and ensuring access to clean and affordable energy (SDG7) are still behind track, and the gaps are not yet ready to be rapidly fulfilled. As part of Goal 7, energy consumption-related targets still lack the potential to be targeted. Considering these elements, this study aims to determine the impact of poverty and gender equality on the energy consumption of households in several European countries. Using data from 2010 to 2022 and the moments’ quantile regression method combined with a pooled OLS based on Driskoll-Kraay estimators, we found statistically significant results regarding the impact of poverty and gender on household energy consumption. These findings’ significance will direct policy design towards those meaningful tools that will increase energy efficiency, address energy poverty, and ensure energy just transition, leaving no one behind. Based on the main findings, the policymakers can understand that a mix of policies is significantly more efficient. In such circumstances, social and economic inequalities will not successfully address development issues without including key environmental priorities, such as emissions mitigation and energy consumption patterns.

1. Introduction

A fundamental problem of the 21st century, the shift to a sustainable energy system, has important ramifications for the economy, social justice, and ecology. This is why the latest COP29 summit emphasized how important it is to help less developed nations with their energy transition. Implementing clean energy policy and climate resilience measures in these nations requires at least USD 300 billion annually. However, most countries with high poverty levels are also known for high corruption and institutional weaknesses. In such circumstances, it is difficult to assess the success of these financing resources. For example, countries such as Guinea-Bissau, Honduras, Yemen, Somalia, Afghanistan, Haiti, Burundi, and South Sudan will need significant corruption mitigation before fully benefiting from the financial support package, struggling simultaneously with chronic poverty and limited institutional prowess.

Previously, the nexus between poverty and energy consumption patterns was described by the energy paradox [1]. Specifically, poor households pay more for their energy consumption than richer people do, only because they cannot afford to invest in energy-efficient goods or more efficient fuels. Under these conditions, they are likely to remain poor [2]. Additionally, energy poverty will lead to indoor pollution, added carbon emissions, biodiversity loss, and health problems [3]. Additionally, environmental degradation is influenced not only by energy consumption but also by dirty sources. Climate change also influences energy consumption patterns as energy costs increase to alleviate the summer conditions with increased temperatures. These issues will further create economic constraints, deepening the general poverty and pressure on the public budgets, which are supposed to address social issues [4].

In terms of gender inequalities, SDG 5 evokes the significance of gender equality for sustaining a sustainable future for all. Poor communities that are struggling to overcome stringent social and economic challenges are also polarized, losing the benefits of involving all social categories in active citizenship on gender grounds. Previous studies have shown that women are heavily involved in decisions about household energy use [5]. For example, involving women in energy decision-making can result in a 25% reduction in electricity costs [6].

Women have more influence on households’ energy consumption patterns [7]. At the same time, involving women in energy decisions can reduce electricity bills by a quarter [5]. In the meantime, women’s involvement in management reduces carbon emissions [6,8,9,10]. However, the majority of previous research has examined these connections in narrow regional contexts or developing nations, failing to offer a thorough analysis of the interactions among poverty, gender, and energy use in European nations. This study aims to fill this research knowledge gap by examining how gender equality and poverty affect household energy usage in several European nations between 2010 and 2022.

The methodological approach, which combines the method of moments’ quantile regression (MMQR) and pooled OLS based on Driskoll-Kraay estimators, represents a novel element for this study. In this sense, to check the validity of our estimates, after applying the MMQR framework to check the significance of our estimates on different quantiles (25th, 50th, 75th, and 90th), we employed a general mean-based regression model to evaluate how the nature of the sign and the statistical significance remain. Additionally, to enhance the robustness of the results, we used a novel approach, such as the Juodis, Karavias, and Sarafidis Granger non-causality test.

The work hypotheses are:

H1. 

The poverty impacts the energy consumption in European countries;

H2. 

The gender inequalities affect the energy consumption patterns in European countries;

H3. 

Public spending for greening purposes improves the environmental quality.

Thus, this study distinguishes itself from prior research through its methodological approach and the geographical scope of its analysis. Recent studies on energy poverty in emerging economies have highlighted the direct impact of female labor market participation on household energy efficiency [11,12]. Conversely, research focusing on developed countries emphasizes that the transition to cleaner energy sources is often influenced by socio-economic factors as well as public support policies [13]. A prior study on European countries also demonstrated that women in parliamentary positions play a significant role in promoting energy efficiency initiatives, thus supporting a sustainable energy transition [14].

Therefore, this study provides a new perspective on the interaction between poverty, gender, and energy consumption in European countries and offers relevant policy recommendations to ensure a just and equitable energy transition.

The outline of this manuscript is as follows. The next section reviews the most significant previous research, while section three presents the methods and data selection. Part four is dedicated to presenting the main results and discussing the study’s findings. The last section concludes the research and sets several policy recommendations.

2. Literature Review

Energy consumption patterns, including the quantity and the energy sources, are responsible for environmental degradation [15,16]. On one hand, the low level of energy efficiency contributes to adding more carbon emissions and environmental degradation [17]. On the other hand, the availability of and access to energy produced from clean sources is also a significant factor in creating environmental pressures. In 2022, more than 41 million European citizens were struggling to warm their homes, according to European Parliament reports. Energy poverty is a cause and an effect of general poverty, being driven by the low income of the households and low energy efficiency of buildings and appliances, creating a cycle of causes and effects in which households reduce their energy consumption without improving the quality of it, negatively affecting the health and well-being of the household’s members.

Energy poverty affects women’s health and safety due to the living conditions and daily chores that expose them to the harmful effects of dirty energy sources or in procuring the fuels for their homes (e.g., collecting wood and other agricultural residuals). In such circumstances, women continue to be less involved in the labor market or access education. The cycle of causes and effects involving energy consumption patterns continues to deteriorate their social welfare and economic status, creating the premises for further environmental degradation [18].

In this sense, energy poverty and gender inequalities were debated in numerous studies.

2.1. Poverty vs. Energy Consumption Patterns

Most of the previous research orbits around the concept of energy poverty, explaining that this status pushes households to use solid fuels instead of electricity [19], while governmental subsidies are still far from being efficient [20]. As a paradox, it was discovered that most of the poor households consume as much energy as economically secure ones due to their buildings’ low-efficiency rates and energy-intensive appliances [21], deepening their economic instability. The poverty-energy nexus is affected by the rural-urban asymmetries, especially in developing and poor world economies. Urban households use more electricity, being better connected to national energy grids and having a better potential to use cleaner energy sources that power these grids. On the other hand, rural areas are less connected to standardized energy grids, widely depending on biomass and agriculture residuals for fueling their heating or cooking [22]. Rural areas are traditionally less advantaged in terms of modern energy access and infrastructure.

Additionally, financial development seems to create more room for energy poverty [23], increasing the general poverty cleavage between urban and rural regions. However, in developed countries, financial development promotes environmental issues mitigation [17], which confirms that access to financial resources is necessary but still not enough if the society has not reached a certain development threshold. The reason for this seems apparent if we are thinking about how poor rural areas are disconnected from modern facilities, which demands increased investment efforts to close the development gaps between the two environments in developing countries. Under these circumstances, energy poverty will remain a significant challenge even if clean energy sources start claiming a bigger share of the total energy supply. A recent study demonstrated that renewable energy has limited to no contribution to mitigating energy poverty in a context characterized by income inequality [24], and low-income households have difficulties in embracing better energy-efficient conditions and clean energy sources.

2.2. Gender Representation vs Energy Use

Numerous previous studies investigated gender-driven energy consumption patterns in various regional contexts.

Socially unequal representation of women, educational gaps, and poverty go hand in hand when it comes to keeping a society developing; consider scholars [25]. Obviously, energy poverty earmarks marginalized social groups or regions, deprived of basic commodities such as access to formalized energy sources, challenged by financial constraints, and forced to gather supplies for their basic needs such as fuels for heating or cooking from informal sources. Energy poverty and energy vulnerability remain a matter of income level, labor market status, and gender in many cases [26]. Moreover, when it comes to adopting cleaner energy or investing in energy-efficient appliances, access to information and the trust in it provided by a certain level of education are critical features of success [27]. Previous research discovered significant differences between men and women in energy use for developed European economies [28]. However, much has changed in the last decade in terms of a general way of living, and certain energy patterns need to be updated. Additionally, demographics have changed due to recent immigration waves and population aging, with Western developed economies also being affected by climate change, modifying energy consumption patterns [29]. One example of pattern alteration in recent years is that while households’ energy consumption concentrated on heating and cooking purposes ten years ago, recent studies approach the consumption behaviors related to cooling the residential spaces [30]. Additionally, starting from a previous study [31], which documented the effects of women’s involvement in policy shaping for electricity access and energy efficiency in developing countries, this research includes features related to women’s involvement in the decision-making process; this time for a different world region with different social and economic characteristics.

Additionally, this approach, this study aims to discuss the impact of gender diversification on energy consumption in decision-making.

3. Material and Methods

3.1. Data

This study utilizes four variables (Poverty ratio, Female labor force participation, Women in national parliament, and Public environmental spending) to explain their impact on the household’s energy use (as the dependent variable). We have started from the Sustainable Development Report (SDR) data and selected only countries for which data were available for the analyzed period. All the variables are presented in

Table 1. Description of study’s variables.

Variables	Abbrs.	Def.	Sources
Households’ energy use	HEU	Final consumption households’ energy use	Eurostat [33]
Poverty ratio	POV	Poverty headcount ratio at USD 3.65/day (2017 PPP, %)	SDR 2024 [34]
Female labor force participation	FMLAB	Ratio of female-to-male labor force participation rate (%)	SDR 2024 [34]
Women in national parliament	WPAR	Seats held by women in national parliament (%)	SDR 2024 [34]
Public environmental spending	PSEP	Public environmental spending (million euro)	Eurostat [35]

. Figure 1 shows the evolution of the variables over time in the 9 European countries analyzed, for which the variables included in the model were covered by data. All data were transformed using natural logarithms and were analyzed using Stata 18. The panel includes nine countries during the 2010–2022 period. Our study covered this period because of the lack of available data regarding households’ final energy consumption for a longer period. The objective of this study was to determine the impact of poverty and gender on households’ energy use. Accordingly, we have selected the poverty headcount ratio at USD 3.65/day and not USD 2.15/day because the selected countries are high-income countries and not low-income countries for which USD 2.15/day would have been a more reasonable indicator, as specified by the World Bank glossary [32]. Moreover, to show the gender impact, we have selected Female labor force participation and Women in the national parliament to highlight both the active involvement of females in the labor market and the presence of women at the highest level of the decision-making process. In this sense, the study considered both the individual and collective impact of women’s actions.

The methodological approach of the paper is highlighted in Figure 2 and is based on several steps, including presenting the descriptive statistics of our variables, performing preliminary tests (correlation variables, cross-sectional dependency test), the evaluation of the stationarity of the variables, the cointegration test, model estimation, exploring the causal relationships between our variables, and a robustness check.

3.2. Regression Estimation and Robustness Testing

To evaluate the impact of the independent variable on the dependent variable of our model, we use two methods: the Method of Moments Quantile Regression (MMQR) and Pooled OLS based on Driskoll-Kraay estimators to check the robustness of our analysis.

The general model is given by the equation below Equation (1):

$${\mathrm{H}\mathrm{E}\mathrm{U}}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{P}\mathrm{O}\mathrm{V}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{F}\mathrm{M}\mathrm{L}\mathrm{A}\mathrm{B}}_{\mathrm{i}\mathrm{t}}+{{\mathsf{\beta }}_3{\mathrm{W}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{i}\mathrm{t}}+\mathsf{\beta }}_4{\mathrm{P}\mathrm{S}\mathrm{E}\mathrm{P}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

3.3. The Method of Moments Quantile Regression (MMQR)

The Method of Moments Quantile Regression (MMQR) was used to increase the reliability of our estimators, which was developed by applying the Pooled OLS based on Driskoll-Kraay estimators. In this sense, using the [36] approach, the results showcased the impact of the independent variables on the dependent variable of our model in the 25th, 50th, 75th, and 90th quantiles. The general model can be described by the following Equation (2):

$$y_i=x_i{\beta }_q+e_i$$

(2)

where ${\beta }_q$ is the vector of unknown parameters associated with the $q^{th}$ quantile. The $e_i$ represents the model prediction error.

3.4. Pooled OLS Based on Driskoll-Kraay Estimators

For testing the robustness of our results, this study used Driscoll and Kraay’s [37] standard errors for coefficients estimated by pooled OLS.; in this sense, the estimation procedure assumes that the error structure is heteroscedastic, autocorrelated, and possibly correlated between the groups (panels) [38].

Considering the form of a linear regression model (Equations (3)–(8)),

$$y_{it}=x_{it}^{\prime }\theta +{\epsilon }_{it}$$

(3)

where i = 1, …, N and t = 1, …, T, x_it = a (K + 1) × 1 vector of independent variables, and θ is a (K + 1) × 1 vector of unknown coefficients.

$$\widehat{\theta }={{(X}^{\prime }X)}^{-1}{X}^{\prime }y$$

(4)

Then, Driscoll and Kraay’s standard errors will take the form:

$$V\left(\widehat{\theta }\right)={\left(X^{\prime }X\right)}^{-1}{\widehat{S}}_T{\left(X^{\prime }X\right)}^{-1}$$

(5)

where

$${\widehat{S}}_T={\widehat{\mathsf{\Omega }}}_0+\sum _{j=1}^{m\left(T\right)}w\left(j,m\right)[{\widehat{\mathsf{\Omega }}}_j+{\widehat{\mathsf{\Omega }}}_j^{\prime }]$$

(6)

With m (T) the lag length up to which the residuals may be autocorrelated and

$${\widehat{\mathsf{\Omega }}}_j=\sum _{t=j+1}^T{h}_t\left(\widehat{\theta }\right)h_{t-j}{\left(\widehat{\theta }\right)}^{\prime }$$

(7)

With

$$h_t\left(\widehat{\theta }\right)=\sum _{i=1}^{N\left(t\right)}{h}_{it}\left(\widehat{\theta }\right)$$

(8)

3.5. Calusality Test

To enhance our results’ relevance, we used the method developed by Juodis, Karavias, and Sarafidis in 2021 [11]. The test reports the statistics and p-value of the Wald test, the hypothesis (null and alternative), and the Half-Panel Jackknive (HPJ) bias-corrected pooled estimator. The Bayesian information criterion (BIC) also helps determine the lag-length selection. Cross-sectional dependence and cross-sectional heteroskedasticity in errors are allowed with a balanced panel condition.

In this sense, according to [39], starting from linear dynamic panel data (Equation (9))

$$y_{i,t}={\mathrm{\varnothing }}_{0,i}+{\sum }_{p=1}^P{\mathrm{\varnothing }}_{p,i}{y}_{i,t-p}+{\sum }_{p=1}^P{\beta }_{p,i}{x}_{i,t-p}+{\epsilon }_{i,t}$$

(9)

With, i = 1, …, N; t = 1, …, T; $x_{i,t}$ a scalar, ${\varnothing }_{0,i}$ as individual-specific effects; ${\epsilon }_{i,t}$ are errors; ${\varnothing }_{p,i}$ as heterogeneous autoregressive coefficients; ${\beta }_{p,i}$ as heterogeneous feedback coefficients (Granger causality parameters).

The HPJ estimator is given starting from the [40] estimator and is bias-corrected further (Equation (10))

$$\tilde{\beta }=\widehat{\beta }+\left\{\widehat{\beta }-{\displaystyle \frac{1}{2}}\left({\widehat{\beta }}_{1/2}+{\widehat{\beta }}_{2/1}\right)\right\}=\widehat{\beta }+T^{-1}\widehat{B}$$

(10)

where ${\widehat{\beta }}_{1/2}$ represents the estimator for the first half of observations and ${\widehat{\beta }}_{2/1}$ refers to the second half of observations.

4. Results and Discussion

4.1. Descriptive Statistics and Correlation Between the Time Series

Table 2. Descriptive statistics of the variable.

	HEU	POV	FMLAB	WPAR	PSEP
Mean	648,572	0.6727436	82.64331	31.34872	7354.444
Maximum	2,671,611	1.919	87.73	44	28612
Minimum	43,783.83	0.097	74.637	14.4	194.7
Std. Dev.	802,785.8	0.4173035	2.614161	7.386143	8234.116
Skewness	1.335784	1.059323	−0.6185612	−0.3600345	0.8078276
Kurtosis	3.241434	3.572492	3.106539	2.157388	2.272011
Observations	117	117	117	117	117

Note: HEU = Households’ energy use; POV = Poverty ratio; FMLAB = Female labor force participation; WPAR = Women in national parliament; PSEP = Public environmental spending.

highlights the descriptive statistics for the analyzed variables. Figure 3, Figure 4, Figure 5 and Figure 6 provide hexagon plots for variable interaction. We can observe that households’ energy use has evolved from 43 thousand (Slovenia) to 2.6 million (Germany). Additionally, the poverty ratio had an oscillatory evolution but mainly increased over the years, presenting a minimum value of 0.097 in Germany and a maximum value of 1.919 in Bulgaria. Moreover, the female-to-male labor force participation rate fluctuated over time from 74% to 87%, while the seats held by women in the national parliament fluctuated from 14% to 44% over the panel. Moreover, kurtosis and skewness were used to assess the normality of the data.

Additionally,

Table 3. Pairwise relationships between variables.

	HEU	POV	FMLAB	WPAR	PSEP
HEU	1		.
POV	0.256 **	1
FMLAB	0.162 **	0.048 **	1
WPAR	0.063 **	0.170 **	0.377 **	1
PSEP	0.227 **	0.322 **	0.504 **	0.582 **	1

Note: ** denotes significance levels of 5%.

suggests no strong association between the employed variables, as the correlation coefficients are all less than 0.7.

4.2. Estimation of the POV, FMLAB, WPAR, and PSEP on HEU and Discussion

Table 4. Testing for stationarity using the Levin, Lin & Chu test.

	At Level		At First-Difference
	Statistic	Prob.	Statistic	Prob.
HEU	−3.5257 ***	0.0002	−8.1565 ***	0.0000
POV	0.6909	0.7552	−4.9207 ***	0.0000
FMLAB	−1.7963 **	0.0362	−2.0604 ***	0.0197
WPAR	−2.7203 ***	0.0033	−4.8766 ***	0.0000
PSEP	4.6094	1.0000	−1.7678 **	0.0385

Note: HEU = Households’ energy use; POV = Poverty ratio; FMLAB = Female labor force participation; WPAR = Women in national parliament; PSEP = Public environmental spending. *** and ** denote significance levels of 1% and 5%, respectively.

showcases the appraisal of the stationarity of the variable. Thus, it can be observed that at the level, the variables analyzed were not stationary, but they became stationary at the first difference.

Cointegration tests such as Kao, Pedroni, and Westerlund were used to evaluate the long-run link between the models’ variables. In this sense, the estimates highlighted a long-run equilibrium relationship and the fact that all panels, or some (for the Westerlung test), are cointegrated (No cointegration hypothesis being rejected). The results of the relationship between the series are given in

Table 5. Testing the cointegration relationship between variables.

Kao Test for Cointegration
	Statistic	Prob.
Modified Dickey-Fuller t	−1.1782	0.1194
Dickey-Fuller t	−3.1188 ***	0.0009
Augmented Dickey-Fuller t	−2.5113 ***	0.0060
Unadjusted modified Dickey-Fuller t	−5.8578 ***	0.0000
Unadjusted Dickey-Fuller t	−5.3652 ***	0.0000
Pedroni test for cointegration
	Statistic	Prob.
Modified Phillips–Perron t	2.3693 ***	0.0089
Phillips–Perron t	−10.4097 ***	0.0000
Augmented Dickey-Fuller t	−8.9172 ***	0.0060
Westerlund test for cointegration
	Statistic	Prob.
Variance ratio	−1.4565 *	0.0726

Note: *** and * denote significance levels of 1% and 10%, respectively.

.

The findings of the cross-section dependence test [41] are presented in Table 3. The results from Friedman and Pesaran tests indicate that the null hypothesis of cross-sectional independence is rejected. This suggests significant evidence of cross-section correlation in the error terms among the countries. The findings are presented in

Table 6. Cross-sectional dependency.

Test Type	Value	Prob.
Friedman	22.125	0.0047
Pesaran	3.182	0.0015

.

Table 7. The impact of POV, FMLAB, WPAR, and PSEP on HEU using the Method of Moments Quantile Regression.

cc	Coefficient	Std. Err.	z	p > |z|	[95% Conf. Interval]
		Quantile regressions for 25th quantiles
POV	−0.2873613 ***	0.0565212	−5.08	0.000	−0.3981408	−0.1765818
FMLAB	−6.330258 ***	1.16067	−5.45	0.000	−8.605129	−4.055388
WPAR	−0.3143962 *	0.1721936	−1.83	0.068	−0.6518894	0.0230971
PSEP	0.727404 **	0.0268455	27.10	0.000	0.6747878	0.7800202
Constant	35.51228
		Quantile regressions for 50th quantiles
POV	−0.2894598 ***	0.0572642	−5.05	0.000	−0.4016956	−0.177224
FMLAB	−5.466089 ***	1.176577	−4.65	0.000	−7.772137	−3.160042
WPAR	−0.3982412 **	0.174483	−2.28	0.022	−0.7402216	−0.0562607
PSEP	0.7412211 ***	0.0272038	27.25	0.000	0.6879027	0.7945395
Constant	32.06087
		Quantile regressions for 75th quantiles
POV	−0.2926665 ***	0.0726811	−4.03	0.000	−0.4351188	−0.1502142
FMLAB	−4.145547 ***	1.493818	−2.78	0.006	−7.073377	−1.217717
WPAR	−0.5263653 ***	0.2215099	−2.38	0.017	−0.9605167	−0.092214
PSEP	0.762335 ***	0.0345355	22.07	0.000	0.6946467	0.8300234
Constant	26.78674
		Quantile regressions for 90th quantiles
POV	−0.2952424 ***	0.092366	−3.20	0.001	−0.4762765	−0.1142082
FMLAB	−3.084786 *	1.866705	−1.65	0.098	−6.743461	0.5738883
WPAR	−0.6292844 **	0.279242	−2.25	0.024	−1.176589	−0.0819802
PSEP	0.7792954 ***	0.0435098	17.91	0.000	0.6940179	0.864573
Constant	22.55016

Note: POV = Poverty ratio; FMLAB = Female labor force participation; WPAR = Women in national parliament; PSEP = Public environmental spending. ***, ** and * denote significance levels of 1 %, 5 %, and 10 %, respectively.

presents the estimations using the method of moment-quantile regression (MMQR) [42]. The findings highlight an indirect relationship between POV, FMLAB, and WPAR with HEU and a direct relationship between PSEP and HEU.

Household energy consumption is important in greenhouse gas emissions [29,43]. According to statistics, in 2022, households represented 25.8% of final energy consumption at the European Union level (the lowest level in the last 6 years). Additionally, the main use of energy by the residential sector [33] was for heating their homes (63.5% of final energy consumption).

The POV estimates disclose a significant and increasing indirect impact from the lower quantile (25th) to the upper (90th) quantile, confirming the study’s first hypothesis. This highlights that an increase in poverty will affect household energy consumption by supporting people’s transfer from centralized energy sources to other individual energy sources.

Bearing in mind that the time spent at home affects the level of energy consumption, our results confirmed the positive role of women’s involvement in the labor force on the consumption of energy by households. According to our results, FMLAB estimates highlight a negative impact on HEC. One should also observe that these coefficients slightly decrease over quantiles while keeping the negative sign. This could explain the possible fall of the variable significance at superior levels of HEC for our model, which represents a limitation of the study. Thus, the effect of FMLAB is stronger when the dependent variable is lower, resulting in an attenuation of the impact when households’ consumption is extensive.

Moreover, considering the mean values of the Pooled OLS based on Driskoll-Kraay (Table 9) results, a unit increase in women’s involvement in the labor force will reduce energy consumption at the household level by 4.88%, while the result is highly statistically significant. This result aligns with [44] in Mozambique and [45] in China. The outcome is explained by the fact that the absence from home limits energy consumption to a minimum (heating for the winter, cooling for the summer, and food preparation). This result is in line with [44], which explained gender roles in household energy consumption patterns, with women being more concerned about the energy for cooking and water heating while having no role in fuel choice decisions.

Furthermore, poverty, as well as the income levels of the households, appears to be important for the household’s energy consumption level. However, our results align with [46], who found a direct relationship between them. The authors showed that when income increases, consumption also increases. In the same direction, a decreasing trend in the poverty rate in analyzed countries will increase household energy consumption. This fact is explained by the automatic or normal transfer of households from using non-centralized types of energy sources (low-tier according to Energy Ladder) to including households in the centralized energy consumption system, which can be more effectively adapted to cleaner and more efficient solutions.

According to the second hypothesis, another relevant aspect of our results (noticeable in Table 7 and Table 9) is linked to the role of women occupying a place in the parliament, which will decrease energy consumption. The result is in line with [46] but different from [44], who consider that the decision-making is equally shared between genders. The MMQR approach confirms the importance of empowering women and their positive role in enhancing awareness and influencing the generalization of responsible behaviors in society. The coefficients are becoming more significant as the dependent variable level is higher.

The involvement of women in the policy-making process improves household energy consumption and could positively influence energy efficiency and conservation. Thus, the aspect of gender equality and inclusivity has become more and more relevant and should be considered in the following policy adjustments [47,48].

Although the literature found a positive effect of public environmental policies and spending on green growth [49,50], our results indicate a direct relationship between public expenditure for environmental protection and HEC. The coefficients are statistically significant over the quantiles, confirming the third hypothesis. However, the results are not in line with the findings of [51] concerning the impact of this type of public spending on carbon emissions. However, when evaluating the impact of environmental R&D expenditure, they also find a negative impact on the environment.

4.3. Causality Test

To test the causal relationship between the variables, we applied the Juodis, Karavias, and Sarafidis Granger non-causality test, which outperforms other test results [39].

Table 8. Testing the causality between variables.

H0	HPJ Wald Test	Prob.	Half-Panel Jackknife Estimator	Prob.
lnPOV does not Granger-cause lnHEU	1.7705	0.1833	0.0216492	0.183
lnFMLAB does not Granger-cause lnHEU	2.6188	0.1056	−0.5990932	0.106
lnWPAR does not Granger-cause lnHEU	0.0918	0.7619	0.0093636	0.762
lnPSEP does not Granger-cause lnHEU	0.1586	0.6905	−0.0145545	0.690
lnHEU does not Granger-cause lnPOV	1.3584	0.2438	−0.7267441	0.244
lnFMLAB does not Granger-cause lnPOV	0.0251	0.8740	0.2784988	0.874
lnWPAR does not Granger-cause lnPOV	4.6538 **	0.0310	−0.2940391	0.031
lnPSEP does not Granger-cause lnPOV	1.4477	0.2289	0.2343152	0.229
lnPSEP does not Granger-cause lnFMLAB	6.8005 ***	0.0091	0.0169243	0.009
lnPOV does not Granger-cause lnFMLAB	6.7273 ***	0.0095	−0.0091481	0.009
lnHEU does not Granger-cause lnFMLAB	9.7250 ***	0.0018	−0.0491751	0.002
lnWPAR does not Granger-cause lnFMLAB	3.4571 *	0.0630	0.0110766	0.063
lnHEU does not Granger-cause lnWPAR	0.1450	0.7034	0.0863737	0.703
lnPOV does not Granger-cause lnWPAR	0.6732	0.4119	0.0322026	0.412
lnFMLAB does not Granger-cause lnWPAR	6.3388 ***	0.0118	−2.454502	0.012
lnPSEP does not Granger-cause lnWPAR	0.0447	0.8326	0.0188011	0.833
lnHEU does not Granger-cause lnPSEP	1.0393	0.3080	−0.2063783	0.308
lnPOV does not Granger-cause lnPSEP	2.3750	0.1233	0.0624745	0.123
lnFMLAB does not Granger-cause lnPSEP	54.4706 ***	0.0000	5.984642	0.000
lnWPAR does not Granger-cause lnPSEP	0.0000	0.9994	0.0000567	0.999

Note: HEU = Households’ energy use; POV = Poverty ratio; FMLAB = Female labor force participation; WPAR = Women in national parliament; PSEP = Public environmental spending. ***, ** and * denote significance levels of 1 %, 5 %, and 10 %, respectively.

shows that the null hypothesis of non-Granger causality is rejected for the lines bolded in the table. These results show that past values of the independent variable contain information that helps predict the dependent variable values. Thus, the results from the JKS Granger test showcased a bidirectional causality between FMLAB and HEU and PSEP—FMLAB and unidirectional causality from WPAR to POV and FMALB, and from POV to FMLAB, from FMLAB to WPAR. This shows that predictive causation can not be found in all relationships between investigated variables.

For the rest of the estimations, one fails to reject the null hypothesis of non-Granger causality. However, some cautions are necessary for the presence of cross-sectional dependence, and when the number of cross-sections is small, this can be noted as a limitation of our study and can be addressed by further research, as [39] already mentioned.

4.4. Robustness Test

The estimation results using Pooled OLS based on Driskoll-Kraay as a robustness test are presented in

Table 9. The impact of POV, FMLAB, WPAR, and PSEP on HEU using Pooled OLS based on Driskoll-Kraay.

Variable	Driscoll-Kraay
POV	−0.2629082 *** (0.026179)
FMLAB	−4.888201 *** (1.281011)
WPAR	−0.4134807 *** (0.15138)
PSEP	0.7442412 *** (0.0223104)
Constant	29.5925 *** (5.145968)
R2	0.921

Note: POV = Poverty ratio; FMLAB = Female labor force participation; WPAR = Women in national parliament; PSEP = Public environmental spending. Standard errors in parentheses *** p < 0.01.

. The findings maintain the results generated by the MMQR technique, pointing out an indirect relationship between POV, FMLAB, and WPAR with HEU and a direct relationship between PSEP and HEU. It can be observed that the impact, sign, and statistical significance of the results are similar to those of the pooled MMQR method.

As can be seen, the Pooled OLS based on the Driskoll-Kraay approach confirms the results of the quantile framework regarding the impact of independent variables on the dependent variable, enhancing the validity of our results. According to our results, a unit increase in POV, FMLAB, and WPAR will reduce energy consumption at the household levels by 0.26%, 4.88%, and 0.41%, respectively, while the result is highly statistically significant. PSEP’s impact on HEC remains statistically significant and positive, overviewing the household’s energy consumption increase supported by specific local and governmental policies to enhance households’ access to centralized energy consumption systems and improve citizens’ social welfare. As supported by [52], public environmental spending is critical for the level of a nation’s environmental and social performance. Thus, by connecting specific environmental targets related to the energy mix, the government can influence the energy consumption potential of the citizens and direct it towards clean energy sources. Taking the first main result, the empirical analysis describes the effects of poverty on households’ energy consumption. Specifically, the financial constraints push households to reduce their energy use to a minimum and direct their resources to other basic needs. However, these results should be understood as a reduced use of standardized energy and, in many cases, limited access to formal sources of energy. Many poor households are not connected to the national grid or formal energy provider. Nevertheless, they use informal sources for heating and cooking needs, such as wood, crop wastes, and other sources that add emissions and indoor pollution, which are responsible for health issues. In many cases, women are the first to suffer, as they are responsible for the errands that are connected to heating and cooking. Moreover, women involved in formal activities are connected to an increased level of education. Theories discuss the inclination of educated people towards a more environmentally-responsible behavior. On the same note, education creates the premises for women’s involvement in policy-making processes and higher representation.

5. Conclusions and Policy Implications

As environmental degradation and biodiversity loss become critical issues for our planet, sustainability encompasses a range of important concerns for decision-makers, including poverty and gender inequality. Additionally, energy security and transition raise complex challenges for macroeconomic stability and public budgets. Increasing the share of renewable energy sources is supposed to result in significant resource allocation from both private and public agents. Significant renewable energy potential in several European countries can diminish the environmental pressure mixed with increased energy efficiency and population awareness.

The practical context of this research reflects the debates on how poverty is still a serious matter in Europe and affects households’ access to centralized energy sources. Additionally, in the European Union, the number of people not being able to heat their homes adequately increased from 6.9% in 2021 to more than 10% in 2023 [53].

This research aims to evaluate the complex connections between those sustainability features mentioned above by understanding the effects on energy consumption patterns.

The study’s main findings confirm that poverty affects the energy consumption in European countries, validating the first hypothesis (H1). At the same time, the second hypothesis (H2) is confirmed as the results document the impact of gender features on energy consumption patterns. The third hypothesis (H3) is validated, confirming that environmental public spending mitigates the ecological degradation. The results of this study are valuable in their empirical contribution and also from a practical standpoint. Starting from these original findings, policymakers can better understand the complexity of targeting sustainability and designing public intervention instruments directed not only to approach a certain goal but also to address various social and economic issues in an integrative manner.

The main results, which are significant for all four quantiles, show the same sign for all the independent variables. At first sight, some of these findings may not envelop the complex meaning of the economic and social realities, while others are intuitive.

In conclusion, while reducing energy consumption due to poverty may seem environmentally beneficial, it often leads to a switch to higher-emission alternatives. In addition, the formal energy sources can benefit from diversification and energy transition towards an increased share of renewables, while the other option is strictly connected with solid fuels and other sources from the energy ladder bottom levels. While the EU invests in renewable sources of energy and goods using such sources, poverty hinders these efforts by not allowing the Europeans to fully benefit from these green solutions. Under these circumstances, the policymakers should consider not only providing the framework for increasing the renewable energy sources but also those economic and social features that still fall behind and keep people from using these green options on a wide scale.

Women’s involvement in the labor market is also connected with decreasing energy consumption. However, this result should be understood differently from the previous. In this case, it cannot be about reducing energy consumption due to the household’s financial constraints. Women participating in the labor market means additional income for the family’s budget, which, theoretically, allows modern comforts such as investing in new appliances and home improvements to increase energy efficiency. Additionally, women’s employment means less time for house activities, which can also be responsible for energy use reduction. As women are more prone, according to various theories, to more environmentally friendly behavior, one can expect from this, policies dedicated to environmental protection and more public resources directed towards protecting environmental integrity.

It is only natural to understand that involving more public resources in environmental protection will achieve ecological sustainability. These research results show a significant correlation between this variable and energy consumption. However, without an insightful explanation of this result, one could understand the opposite. The energy consumption increase effects of environmental public spending must be understood in the sense of investing in more renewable sources of energy and also getting an increased number of households connected to the national grid and other formal sources of energy supply. Certainly, by investing in clean sources of energy and increasing the share of renewables, it targets the environmental issues. However, without having a larger rate of households connected to these clean sources and still depending on informal ways of fueling their daily home activities, they will not achieve sustainability. In this sense, the energy consumption increase effect must be understood as a way of achieving just energy transition and having a larger amount of the population as beneficiaries of clean energy sources.

There are several specific policy recommendations that arise from the study’s results:

–

design mix policies that integrate economic, social, and environmental priorities, allowing people to benefit from technological advancements and enhance their standard of living;

–

design environmental inclusion policies by addressing the marginalized communities by increasing their awareness and capacity to embrace the green European targets;

–

mitigating environmental discrimination by incentivizing marginalized social categories to get actively involved in environmental initiatives;

–

shaping energy consumption behavior by promoting education initiatives

–

offering country-based financing programs for green transition, connected with the nations’ standard of living, average income, social asymmetries, and structural gaps.

Limitations and further directions. One limitation of the study is the indicators used to represent the variables; different poverty lines or headcount ratios could alter the perceived impact of poverty on energy consumption. Future research should explore this area for broader insights and examine the role of active women in household energy consumption using various models. Another limitation of the study consists of the heterogeneous results generated by using the JKS Granger causality test. Further analysis could address this limitation.