Life Expectancy and Its Determinants in Selected European Union (EU) and Non-EU Countries in the Mediterranean Region

Abstract

In the Mediterranean region, countries grapple with a mix of environmental pressures, such as air pollution and climate vulnerability, alongside economic disparities and migration issues. In this context, we aim to highlight the interaction between migration (NMIG), economic growth (GDP), foreign direct investments (FDI), fossil fuel (FF) usage, consumption from renewables (RENC), CO₂ emissions, and life expectancy (LE). This is important for gaining insights into how policies in areas like energy, environment, migration, and FDI influence long-term health outcomes. Our research examines the determinants of LE in two groups of Mediterranean countries (EU-Med8 and Non-EU-Med4) using a panel ARDL approach. The long-run results for Med8 indicate that RENC positively influences LE, while FF has a significant negative effect. Economic growth and migration also play important roles, with GDP positively affecting LE. The error correction term (ECT) confirms convergence toward long-run equilibrium. For Med4, FF consumption and CO₂ negatively affect LE, while migration and FDI exhibit mixed results. These findings suggest that while renewable energy transitions benefit LE in EU Mediterranean countries, challenges persist in non-EU countries, where energy infrastructure and investment patterns may not yet support positive health outcomes.

1. Introduction

1.1. General Context in the Mediterranean Region

In the Mediterranean region, we focus on a selected group of countries divided into the European Union (EU) and non-EU members. Spain, France, Italy, Malta, Slovenia, Croatia, Greece, Cyprus are part of the EU and are located in the Mediterranean Sea. As EU members, they share political, economic, and regulatory frameworks under the EU, and many of them participate in the Euro currency zone and the Schengen area for free movement. Monaco, Albania, Turkey, Montenegro, Bosnia and Herzegovina are in the Mediterranean region but are not part of the EU. They were selected based on their geographic location, availability, and consistency of relevant data, as well as their significant economic or political ties with the EU. However, many have close economic or political ties with the EU. Some of these countries, like Turkey and Montenegro, are (potential) candidates for future EU membership. Investigating the selected Mediterranean countries, comprising both EU and non-EU members, offers an opportunity to explore a region of economic and geopolitical importance [1]. The Mediterranean has long been an essential area for global trade and economic interactions. Today, its strategic significance remains, particularly because it connects Europe, North Africa, and the Middle East, serving as a crossroads for commerce, migration, and security concerns [2,3]. Thus, these groups of countries present several key features that make them valuable for research.

Economically, the Mediterranean is an important hub for international trade and tourism. These countries exhibit diverse economic structures, with advanced economies like France, Italy, and Spain, alongside emerging markets, such as Albania and Bosnia and Herzegovina. Countries like Greece and Turkey, as well as Monaco and Malta, heavily rely on tourism for economic growth [4,5]. However, the Mediterranean’s heavy reliance on tourism and coastal development has led to overdevelopment, which threatens local ecosystems [6]. Countries like Spain and Malta, which depend heavily on tourism, must balance economic growth with the need to protect their natural environments [7]. These dynamics create opportunities for comparative studies between EU and non-EU countries.

From an environmental and sustainability perspective, the Mediterranean faces several pressing challenges, including climate change, water scarcity, and pollution [8,9]. Coastal regions in countries like Greece, Cyprus, Turkey, and Spain are highly vulnerable to rising sea levels and extreme weather events [10,11]. Renewables (RES) development is becoming increasingly important as countries shift away from fossil fuels and seek sustainable solutions [12,13]. This aspect presents a critical research topic on how Mediterranean countries, both EU and non-EU, are addressing climate risks, adopting RES policies, and managing resources like water [14,15].

Another pressing issue in the Mediterranean is migration and demographic change. The region is at the forefront of migration flows from Africa and the Middle East into Europe, with countries such as Greece, Italy, and Turkey acting as major entry points [16,17]. These movements create demographic pressures and challenges related to integration and labor market absorption [18]. Non-EU countries like Albania and Montenegro, which have younger populations compared with aging EU countries like Italy and Spain, face distinct demographic challenges, making migration and demographic research essential for understanding labor market dynamics, health issues [19], and population growth in the region [20].

1.2. Life Expectancy in the Mediterranean Region: Interlinking the Input Variables

LE in the Mediterranean region varies significantly depending on the country, influenced by factors such as healthcare quality, lifestyle, economic conditions, and environmental factors [21]. In the EU Mediterranean countries, Spain has one of the highest LEs in the world, averaging around 83 to 84 years. The country’s Mediterranean diet, high-quality healthcare system, and strong social support contribute to this longevity [22]. In the non-EU Mediterranean countries, Monaco stands out with the highest LE globally, reaching around 85 to 86 years, due to excellent healthcare and affluent living conditions. In Bosnia and Herzegovina, LE is slightly lower, at about 76 to 77 years, affected by the quality of healthcare and economic challenges. Overall, Mediterranean countries tend to have higher LE rates compared with the global average, especially in the EU nations. However, non-EU countries generally show slightly lower LE due to healthcare and economic disparities [23].

Studying LE in relation to factors such as NMIG, RENC, FF, GDP, CO₂ emissions, and FDI in the Mediterranean region provides an understanding of how environmental, economic, and social factors interconnect to influence public health and well-being [24]. Migration impacts both the demographic structure and social dynamics of a region, influencing healthcare systems, labor markets, and economic stability. High levels of immigration put pressure on public services, including healthcare, which may affect LE. Conversely, emigration can lead to a loss of skilled healthcare workers and other professionals, potentially lowering the quality of services.

The shift toward RES is directly linked to improvements in environmental quality, particularly air quality. Poor air quality, often from FF, is a major contributor to respiratory and cardiovascular diseases, which affect LE [25]. Countries with higher RENC typically have lower pollution levels, which can result in healthier populations and extended LE. FF dependence is also associated with economic vulnerabilities due to fluctuating oil prices, which may affect a country’s ability to invest in healthcare and social services [26,27]. Elevated levels of CO₂ emissions are an indicator of poor environmental conditions, often stemming from industrial activities, transportation, and FF. High CO₂ emissions contribute to climate change, which has direct and indirect effects on public health through extreme weather events, food security issues [28], and heat-related illnesses. FDI influences economic growth, job creation, and infrastructure development, including healthcare infrastructure. When directed toward healthcare, technology, and RES, FDI may help improve living conditions and healthcare access, thereby positively impacting LE [29,30]. On the other hand, if FDI prioritizes high-polluting industries, it may worsen environmental conditions, which could reduce LE by increasing exposure to harmful pollutants [31,32].

Countries with higher GDP can afford better healthcare systems, including hospitals, clinics, preventive care, and emergency services, which directly affect LE. In Mediterranean countries, the disparity in GDP across nations (e.g., higher in EU countries like France and Italy and lower in non-EU countries like Albania or Bosnia) correlates with differences in healthcare quality and accessibility [33]. Wealthier countries with higher GDP often have more resources to invest in environmental sustainability, such as RES projects and pollution control [34,35]. GDP also influences migration patterns. Countries with higher GDP often attract more immigrants seeking better economic opportunities and improved quality of life.

1.3. Novelty, Motivation for Research, and Research Questions

In the Mediterranean region, where countries face a combination of environmental stressors (e.g., air pollution, climate vulnerability), economic disparities, and migration challenges, understanding how these factors interplay with LE is important for humanity. It provides insights into how policy interventions in energy, environment, migration, and foreign investment can shape long-term health outcomes.

This paper makes a contribution by employing a panel ARDL model to analyze the determinants of LE across two groups of Mediterranean countries (EU-Med8 and Non-EU-Med4). While previous studies have examined the relationship between LE and macro-level variables such as GDP, FDI, pollution, or health expenditure, few have focused specifically on the Mediterranean basin, a region marked by significant heterogeneity in environmental stress, energy infrastructure, migration patterns, and institutional capacity. Our work stands out methodologically by applying the panel ARDL framework to these two subgroups separately, allowing us to uncover differentiated short-run and long-run dynamics. This dual-block approach reveals asymmetries in how economic and environmental factors influence LE, which are often obscured in more aggregated models.

This study evaluates the long-run and short-run impacts of RENC, FF, GDP, NMIG, FDI, and CO₂ on LE. The results indicate that, in the long run, RENC positively affects LE in Med8, while FF consumption has a negative impact in both Med8 and Med4. Moreover, migration and FDI show mixed effects across the two groups.

Our research is novel because it addresses a critical and underexplored intersection of environmental, economic, and social factors, namely migration, economic growth, FDI, fossil fuels, renewables, CO₂ emissions, and LE, across a unique and diverse group of Mediterranean countries. By focusing on both EU and non-EU Mediterranean nations, the research captures the disparities in development, environmental pressures, and demographic changes, offering a comparative analysis that highlights the distinct challenges faced by each group. The use of a panel ARDL model to investigate long-run and short-run relationships between these variables and LE in two country groups (EU-Med8 and Non-EU-Med4) brings a novel econometric approach to a region often overlooked in such depth. Additionally, it emphasizes the roles of energy sources, namely RENC and FF, in influencing public health outcomes. This integrative approach, linking energy policies, migration dynamics, and health outcomes, offers new insights into how policy interventions in these areas may improve LE, particularly in regions facing environmental, economic, and social challenges.

Even if variables such as education, urbanization, and health expenditure are frequently examined in the literature on life expectancy, we deliberately did not include them in our baseline model. These variables, while empirically significant, have been extensively studied across many regional and global contexts, and their inclusion may yield expected results that offer limited additional insight. Instead, we focused on less commonly combined but highly policy-relevant variables in the Mediterranean context, such as RENC, FF, CO₂ emissions, GDP, FDI, and NMIG, whose joint effects on LE remain underexplored, particularly in a comparative EU versus non-EU framework.

Therefore, the current research aims to answer the following research questions (RQ):

RQ1: How do energy source transitions (from FF to RENC) affect long-term public health outcomes, as measured by LE, in Mediterranean countries?

RQ2: What is the role of foreign direct investment in shaping health and environmental conditions in EU and non-EU Mediterranean countries?

RQ3: To what extent does economic growth contribute to improving LE, and how is this relationship moderated by environmental stressors (e.g., CO₂ emissions, FF dependence)?

RQ4: How does migration affect the availability and quality of health-related services, and in turn, LE in Mediterranean countries with varying institutional capacities?

2. Literature Review

The existing relevant literature on LE has evolved along three main lines: theoretical foundations linking economic and environmental variables to health outcomes; methodological innovations in panel and time series modeling; and empirical studies assessing determinants of LE across countries and regions. In this section, we organize the review accordingly to clarify the contributions of each strand.

The impact of economic growth, energy consumption, and urbanization (URB) on CO₂ emissions in developing Mediterranean countries was analyzed (1995–2016) [32]. Using a STIRPAT model and panel data analysis, the research highlighted that while GDP and FF consumption increased carbon emissions, URB and RES usage helped reduce them. Another research examined the factors influencing LE in Eastern Mediterranean countries by analyzing socioeconomic and expenditure variables (1995–2006) [36]. Using data from WHO and the CIA World Factbook, a random-effects regression model assessed the impact of 11 covariates on LE. Significant factors included GDP, death rate, infant mortality rate, total fertility rate, female LE, incidence of tuberculosis, and time. Between-country differences were significant for infant mortality and female LE, while within-country effects showed significance for death rate, fertility, and disease incidence. Notably, health expenditure had no significant association with LE, suggesting that demographic and disease-related factors were the main predictors of LE in this region.

The influence of RES, URB, pollution, and GDP on LE in Somalia (1990–2016), utilizing the Kernel regularized least square method, was investigated [37]. The findings revealed that RES, GDP, and URB positively impacted LE, while pollution was statistically insignificant. The research also identified bidirectional causality between RES and LE, GDP and LE, and pollution and LE, with unidirectional causality from URB to LE. Additionally, data from 21 Eastern Mediterranean Region (EMR) (1995–2010) were analyzed [24] to identify factors influencing LE. Using cluster analysis, countries were grouped, and a multilevel model predicted LE based on variables like GDP, vaccination rates, and URB. Results show that non-industrialized nations had lower LE, with vaccination rates as a key predictor across all clusters. Industrialized nations had additional positive predictors (GDP and health expenditures). Non-industrialized countries had LE over 14% lower. A health production function also for the EMR using the Grossman model was estimated [38]. Panel data from 21 EMR countries (1995–2007) were analyzed with a fixed-effect model to explore the relationship between LE and socioeconomic factors. Results indicate that income per capita, education, food availability, URB, and employment were significant determinants of LE.

The macroeconomic environment, particularly unemployment, significantly affected health outcomes [39]. The research examined the impact of economic crises on health indicators in Eastern Mediterranean countries, using LE and infant mortality rate as measures of health status and unemployment rate as a proxy for economic crises. Panel data (2005–2014) was analyzed, revealing that a 1% rise in unemployment decreased LE by 0.17 years, while a 1% increase in health expenditure per capita reduced infant mortality by 4.54. Another research analyzed LE in 31 of the world’s most polluted countries (2000–2017), focusing on environmental degradation [40]. Using the Preston curve model and Granger causality tests, the results show GDP positively impacted LE, while environmental degradation, particularly carbon emissions, reduced it. Health expenditure, clean water, and sanitation improved LE.

Usually, LE reflects healthcare system performance, and regional disparities in Indonesia posed a public health challenge [41]. The research identified factors influencing LE and grouped provinces to foster cooperation. Using 2015 data from Indonesia’s Ministry of Health, structural equation modeling (SEM) highlighted key determinants, including education, per capita expenditure, health workforce, and facilities, with expenditure per capita being the strongest factor. Cluster analysis grouped the 34 provinces into five clusters, suggesting that economic improvements were the most effective way to enhance LE. Moreover, the relationship between LE and social determinants in 75 municipalities in Sergipe, Brazil, was analyzed using spatial autocorrelation [42]. The southeastern region showed clusters with better living conditions, while the northwestern and far-eastern areas faced poor conditions. Factors like high dependency ratios, illiteracy, and unemployment among the elderly negatively affected LE. The study confirmed a correlation between poor social indicators and lower LE. Another study examined how publicly funded healthcare and social determinants of health (SDOH) impacted LE across 196 countries and 4 territories [43]. Countries with publicly funded healthcare had significantly higher LEs (76.7 years) compared with those without it (66.8 years). Longer LE was consistently linked to publicly funded healthcare across various SDOH, though disparities remained based on the burden of these factors. Both publicly funded healthcare and reduced inequalities in SDOH were found to be key drivers of longer LE.

The SDOH, specifically LE and infant mortality, were further studied in Ukraine and Poland, focusing on regional differences [44]. Data from Ukrainian and Polish statistical sources were used, including medical statistics, state services, and statistical management bodies, with system analysis, bibliographic, statistical, and analytical methods applied. Over 28 years, both countries saw increases in average LE for men and women, with women consistently living longer. Another research analyzed the key determinants of LE at birth in twelve Southeastern European countries from 2000 to 2015, using a cointegrated panel regression model [45]. The analysis includes Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Cyprus, Greece, Macedonia, Moldova, Romania, Serbia, Slovenia, and Turkey. It examined the impact of six variables: consumer price index (CPI), employment, food production index, gross national income (GNI), health expenditure, and immunization. The results show that CPI and GNI positively influenced LE, while employment had a negative effect. Furthermore, the socioeconomic factors influencing LE in Southeastern Europe (2000–2019) were analyzed [46], using a fixed-effects panel data model. Key findings indicated that GDP, URB, emissions, and fertility rate were significant determinants of LE, while healthcare expenditure, both public and out-of-pocket, had no significant impact. Surprisingly, marital status was found to negatively affect LE in one of the regressions.

Some research examined factors influencing LE in 136 countries from 2002 to 2010, focusing on social, economic, and environmental determinants [47]. Using panel data analysis, the results show that unemployment and inflation negatively impacted LE, while gross capital formation and GNI had positive effects. URB was identified as a key socioenvironmental factor contributing to mortality. Also, the socioeconomic determinants of LE in Nigeria from 1980 to 2011, using VAR and VECM models, were investigated [48]. Key variables included secondary school enrollment, health expenditure, income, unemployment, and the Naira exchange rate. Contrary to expectations, conventional factors like income, education, and health expenditure were not significant predictors of LE in Nigeria. According to [49], Nigeria’s LE was 54.33 years, below the sub-Saharan African average of 61.27 years and the global average of 72.56 years. The determinants of LE in Nigeria from 1981 to 2017 were studied using the autoregressive distributed lag (ARDL) method. In the short run, real GDP, past inflation, imports, and government consumption expenditure positively impacted LE, while current inflation, imports, household consumption, and exchange rate had negative effects. In the long run, real GDP, household consumption, and exchange rate positively influenced LE, while inflation, imports, and government consumption had negative impacts. Another study explored the determinants of health using three metrics: LE, healthy LE (HLE), and the gap between them [50]. Causal machine learning and statistical methods were used to analyze the effects of various socioeconomic factors. Results show that increased access to basic water services and public health expenditure raised LE, while high HIV prevalence and poverty rates lowered it. Higher GNI and moderate BMI increased HLE, but high HIV rates reduced it. The gap between HLE and LE widened with higher public health spending and GNI but narrowed with high HIV rates and moderate BMI.

The socioeconomic determinants of LE in Malaysia using data from 12 states and one federal territory were examined (2002–2014) [51]. Panel data analysis revealed that poverty and income significantly affected LE for both genders, while unemployment impacted female and total LE. Income inequality and public health spending did not show significant effects. Another research examined the factors influencing LE in Pakistan (1975–2020) [52], with a focus on environmental degradation, particularly CO₂ emissions. Long-run results from the ARDL model show that CO₂ emissions, inflation, food production, and death rate negatively affected LE, while income, URB, population growth, birth rate, health expenditure, and education positively impacted it. Furthermore, the relationship between GDP and LE in Pakistan was analyzed [53], considering the roles of financial development and energy consumption. Unit root tests and the ARDL bounds testing technique, accounting for structural breaks, were used to explore cointegration among the variables. The findings confirmed cointegration, with GDP positively linked to LE. However, financial development negatively impacted LE, and energy consumption reduced it through environmental degradation.

LE in five SAARC countries, Bangladesh, India, Pakistan, Nepal, and Sri Lanka, was investigated [54] (2000–2016). Significant positive factors included mean years of schooling and sanitation services, while total fertility rate, URB, and emissions negatively impacted LE. Surprisingly, health expenditure had a negative effect. Other factors like GDP and internet usage were insignificant. Another study aimed to analyze factors influencing HLE [55]. A review of 90 studies (28.9% from China) identified key determinants using an ecological model of health, classifying risk factors into five levels: personal characteristics, behavior and lifestyle, social networks, living/working conditions and socioeconomic, cultural and environmental conditions. Non-communicable diseases and personal characteristics were the most studied factors, reported in 58.9% and 52.3% of studies, respectively. Individual behavior and lifestyle, identified as the most modifiable factors, significantly contributed to global disability-adjusted life years. Moreover, the association between SDOH and LE in China (2005–2020) using provincial-level data was examined [56]. LE increased from 73.1 to 77.7 years during this period, with significant regional disparities, showing high LE clustering in coastal areas and low LE in western regions. Key SDOH factors positively associated with LE included GDP, Gini index, healthcare institution beds, and natural population growth. Another study examined why some countries achieve higher or lower LE than predicted by their income levels [57]. LE relative to GDP was analyzed for all countries, with Ethiopia, Brazil, and the United States examined in detail. Ethiopia exceeded its expected LE by 3 years due to community health strategies, improved water access, female education, and civil society. Brazil exceeded by 2 years, driven by reduced inequality, healthcare access, and political participation. The U.S. fell 2.9 years short, likely due to neoliberal policies, market-based healthcare, and rising inequality.

The literature highlights that LE is influenced by a complex interplay of economic, environmental, demographic, and institutional factors. Theoretically, the Preston curve remains a dominant framework, while recent studies incorporate environmental and energy-related variables. Methodologically, fixed-effects, GMM, and cointegration models are common, with limited applications of panel ARDL for long- and short-run dynamics. Empirically, most studies focus on single countries or broad regional panels, often neglecting the Mediterranean basin as a distinct case. Our contribution is to combine economic (GDP, FDI), environmental (CO₂, RENC, FF), and demographic (NMIG) variables in a unified panel ARDL model, distinguishing between EU (Med8) and non-EU (Med4) countries. This dual-block analysis allows us to detect regional asymmetries and uncover development–health–energy linkages often masked in more aggregated or single-country studies.

A comparative summary of selected studies on LE and related variables is provided in

Table 1. Previous studies on LE comparison.

Ref.	Time Interval	Main Focus/Objectives	Variables	Geographical Scope	Methodology Employed
[32]	1995–2016	Impact of economic growth, energy consumption, and urbanization on CO2 emissions	GDP, fossil fuel consumption, urbanization, renewables	Developing Mediterranean countries	STIRPAT model, panel data analysis
[36]	1995–2006	Factors influencing life expectancy in Eastern Mediterranean countries	GDP, death rate, infant mortality rate, fertility rate, female LE, tuberculosis, time	Eastern Mediterranean	Random-effects regression
[37]	1990–2016	Impact of RES, urbanization, pollution, and GDP on life expectancy	RES, GDP, urbanization, pollution	Somalia	Kernel Regularized Least Squares
[24]	1995–2010	Factors influencing life expectancy using cluster analysis	GDP, vaccination rates, urbanization	Eastern Mediterranean Region (21 countries)	Cluster analysis, multilevel model
[38]	1995–2007	Estimating a health production function for life expectancy	Income per capita, education, food availability, urbanization, employment	Eastern Mediterranean Region (21 countries)	Grossman model, fixed-effects panel regression
[39]	2005–2014	Impact of economic crises on health indicators	LE, infant mortality, unemployment, health expenditure	Eastern Mediterranean countries	Panel data analysis
[40]	2000–2017	Effect of environmental degradation on life expectancy	GDP, CO2‚ emissions, health expenditure, clean water, sanitation	31 most polluted countries	Preston curve model, Granger causality
[41]	2015	Factors influencing LE and regional disparities in Indonesia	Education, per capita expenditure, health workforce, health facilities	Indonesia	Structural equation modeling (SEM), cluster analysis
[42]	2010–2017	Relationship between social determinants and LE in Sergipe, Brazil	Dependency ratios, illiteracy, unemployment among elderly	Sergipe, Brazil (75 municipalities)	Spatial autocorrelation
[43]	Latest available data (2020–2021)	Impact of publicly funded healthcare and SDOH on LE	Public healthcare, SDOH (various)	196 countries and 4 territories	Comparative analysis
[44]	28 years (approx. 1990s–2018)	Regional differences in LE in Ukraine and Poland	Medical statistics, LE by gender	Ukraine and Poland	System analysis, statistical analysis
[45]	2000–2015	Key determinants of LE at birth in Southeastern Europe	CPI, employment, food production index, GNI, health expenditure, immunization	12 Southeastern European countries	Cointegrated panel regression model
[46]	2000–2019	Socioeconomic factors influencing LE in Southeastern Europe	GDP, urbanization, emissions, fertility rate, health expenditure, marital status	Southeastern Europe	Fixed-effects panel data model
[47]	2002–2010	Social, economic, and environmental determinants of LE	Unemployment, inflation, gross capital formation, GNI, urbanization	136 countries	Panel data analysis
[48]	1980–2011	Socioeconomic determinants of LE in Nigeria	Secondary school enrollment, health expenditure, income, unemployment, exchange rate	Nigeria	VAR and VECM models
[49]	1981–2017	Determinants of LE in Nigeria	GDP, inflation, imports, government and household consumption, exchange rate	Nigeria	ARDL model
[50]	LE: 2000–2016 Healthy LE: 2008–2019	Determinants of health using LE, HLE, and their gap	Water services, public health expenditure, HIV prevalence, poverty, GNI, BMI	Global	Causal machine learning, statistical analysis
[51]	2002–2014	Socioeconomic determinants of LE in Malaysia	Poverty, income, unemployment, income inequality, public health spending	Malaysia (12 states, 1 federal territory)	Panel data analysis
[52]	1975–2020	Impact of environmental degradation on LE in Pakistan	CO2 emissions, inflation, food production, death rate, income, URB, birth rate, health expenditure, education	Pakistan	ARDL model
[53]	Not specified	Relationship between GDP and LE in Pakistan	GDP, financial development, energy consumption	Pakistan	Unit root tests, ARDL bounds test with structural breaks
[54]	2000–2016	Determinants of LE in SAARC countries	Schooling, sanitation, fertility rate, URB, emissions, health expenditure, GDP, internet usage	Bangladesh, India, Pakistan, Nepal, Sri Lanka	Panel data analysis
[55]	Databases used for the retrieval of related literatures published by 7 May 2022	Factors influencing healthy life expectancy (HLE)	Personal characteristics, behavior, social networks, working/living conditions, SDOH	Global (90 studies, 28.9% China)	Ecological model, literature review
[56]	2005–2020	Association between SDOH and LE in China	GDP, Gini index, healthcare beds, natural population growth	China (provincial-level)	Regional analysis
[57]	1990–2018	Why some countries exceed or fall short of LE expectations relative to GDP	Community health, water access, female education, healthcare access, inequality	Global (case studies: Ethiopia, Brazil, USA)	Comparative country analysis

.

While existing studies have explored the determinants of LE through various lenses, such as economic growth, environmental degradation, urbanization, and public health spending, no study appears to comprehensively fulfill all the dimensions:

• Integrate multiple dimensions like migration, FDI, energy structure (fossil vs renewables), and CO₂ emissions together as determinants of LE;
• Contrast EU and non-EU Mediterranean countries to reveal regional disparities in long-term health outcomes;
• Apply a dynamic econometric technique such as panel ARDL to disentangle short- and long-run effects of these factors;
• Examine the role of migration in relation to LE in the Mediterranean context, where migration pressure is high and politically salient.

3. Methodology

The input data (1990–2023) were collected from the World Bank and Our World in Data websites. The following econometric analyses are performed on the selected EU and non-EU Mediterranean countries (as in Figure 1): first- and second-generation panel unit root tests (PURT), panel ARDL, and Granger causality.

PURTs are used to examine the existence of unit roots (non-stationarity) in panel data. The first- and second-generation PURTs differ in how they handle cross-sectional dependence, with first-generation PURTs assuming cross-sectional independence, while second-generation PURTs account for potential cross-sectional dependence.

3.1. First-Generation PURTs

Levin, Lin, and Chu (LLC) test by [58] assumes a common unit root process as follows:

$${∆y}_{it}={\alpha }_i+\rho y_{it-1}+{\sum }_{j=1}^{p_i}{\beta }_{ij}{∆y}_{it-j}+{\epsilon }_{it}$$

(1)

where ${∆y}_{it}$ is the first difference of the variable y, ${\alpha }_i$ is the individual-specific fixed effect, ρ is the common autoregressive parameter, ${\beta }_{ij}$ are lag coefficients, and ${\epsilon }_{it}$ is the error term.

The null hypothesis (NH) $H_0$ asserts the presence of a unit root test ($\rho =0$), while the alternative hypothesis (AH) is that the series is stationary ($\rho <0$). The LLC test assumes cross-sectional independence and the common autoregressive parameter ρ.

Im, Pesaran, and Shin (IPS) test by [59] allows for individual-specific autoregressive parameters, meaning that the unit root process can vary across individuals. The IPS test uses the same ADF structure (1), but the statistical interpretation differs.

Each ${\rho }_i$ can differ across individuals. The IPS test averages individual ADF (Augmented Dickey–Fuller) statistics across the panel.

The NH is $H_0:{\rho }_i=0$, for all i (all series have a unit root). AH posits: $H_1:{\rho }_i<0$, for some i (some series are stationary).

The average of individual t-statistics from ADF regressions is

$$W_{IPS}={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{t}_i$$

(2)

where $t_i$ is the ADF test statistic for the i-th unit.

ADF-Fisher test is based on combining individual ADF tests using Fisher’s test [60]. It is non-parametric and does not require the same number of lags for each cross-section. The test statistic is

$$P=-2{\sum }_{i=1}^N\mathrm{l}\mathrm{n}\left(p_i\right)$$

(3)

where $p_i$ is the p-value of the ADF test for the i-th individual. Under the NH, P follows a chi-square distribution with 2N degrees of freedom.

The NH is that all series have a unit root. AH posits that at least one series is stationary.

3.2. Second-Generation Panel Unit Root Tests

The Cross-sectionally augmented IPS (CIPS) test [61] explains the cross-sectional dependence by increasing the standard ADF regression with the cross-sectional averages of the variables and their lags. The model is

$${∆y}_{it}={\alpha }_i+{\rho }_i{y}_{it-1}+{\gamma }_0\overline{y_{t-1}}+{\sum }_{j=1}^{p_i}{\beta }_{ij}{∆y}_{it-j}+{\sum }_{j=0}^{p_i}{\delta }_{ij}\overline{{∆y}_{t-j}}+{\epsilon }_{it}$$

(4)

where $\overline{y_{t-j}}$ is the cross-sectional average of $y_{it}$. ${\gamma }_0$ and ${\delta }_{ij}$ are coefficients capturing the influence of cross-sectional dependence.

The CIPS statistic is computed as the average of individual Cross-sectionally augmented ADF (CADF) statistics:

$$CIPS={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{t}_i(N,T)$$

(5)

where $t_i(N,T)$ is the CADF statistic for the i-th unit.

The NH is that all series have a unit root. The AH posits that at least some series are stationary.

Each test has strengths depending on whether cross-sectional dependence is present and how heterogeneous the panel is in terms of unit root processes.

3.3. Panel ARDL

Given that the variables of interest are a mix of I(0) and I(1) processes, we employ the panel ARDL model, as proposed by [62]. The panel ARDL model can be applied when the variables are I(0) or I(1). This makes it a robust method for analyzing relationships in datasets where variables have mixed integration orders. The panel ARDL model is

$$y_{it}={\alpha }_i+{\sum }_{p=1}^P{\lambda }_p{y}_{it-p}+{\sum }_{q=0}^Q{\beta }_q{x}_{it-q}+{\epsilon }_{it}$$

(6)

where $y_{it}$ is the dependent variable for unit i at time t, ${\alpha }_i$ is the cross-sectional fixed effect (specific to each unit i, allowing for individual heterogeneity), ${\lambda }_p$ is the autoregressive term, ${\beta }_q$ is the distributed lag term. The dependent variable has P lags, while the independent variables have Q lags.

The panel ARDL model allows for heterogeneity by including individual-specific effects ${\alpha }_i$ which captures unobserved heterogeneity across the cross-sections (countries).

Finally, Granger causality tests are used to determine whether one variable can predict another variable, implying a short-term relationship. The idea is that if variations in one variable (X) precede variations in another variable (Y), then X is said to “Granger cause” Y.

Before presenting the estimation results, we briefly address potential endogeneity concerns. In our empirical setting, potential reverse causality may exist, for example, between GDP and LE or between FDI and CO₂ emissions. A wealthier economy can improve public health and thereby increase LE, but a healthier population can also contribute to economic growth. Likewise, FDI may affect environmental quality, but countries with specific environmental characteristics may also attract or repel FDI. These bidirectional dynamics create endogeneity risks that may bias standard estimators. We address this issue by employing a panel ARDL approach. The panel ARDL includes lagged explanatory variables and separates short-run from long-run relationships. This helps reduce simultaneity bias and provides more robust estimates. We also apply Granger causality tests to empirically investigate directional relationships between variables.

It is important to clarify that Granger causality does not imply true causality in a structural sense. Instead, it tests for temporal precedence, indicating whether past values of one variable help predict another. As such, the results should be interpreted as suggestive of predictive directionality, not causal mechanisms.

3.4. Data Quality and Missing Values Considerations

The dataset employed in this study covers the period from 1990 to 2023 and is sourced from reputable databases, including the World Bank and Our World in Data. The extracted time series are complete, with no missing values.

We conducted Breusch–Pagan tests for heteroskedasticity on the log-transformed regression variables. For the Med4 group, the test yielded a p-value of 0.642, indicating no significant heteroskedasticity. For the Med8 group, the test yielded a p-value of 0.1018, indicating no significant heteroskedasticity.

Finally, to check for multicollinearity, correlation matrices for the logged variables are reported in Appendix A (

Table A1. Correlation matrix for Med4.

	LE	GDP	RENC	FF	FDI	CO2	NMIG
LE	1.000	0.384	0.221	−0.186	0.340	0.036	−0.077
GDP	0.384	1.000	−0.219	0.238	−0.362	0.642	−0.591
RENC	0.221	−0.219	1.000	−0.679	0.715	−0.573	0.254
FF	−0.186	0.238	−0.679	1.000	−0.711	0.699	−0.280
FDI	0.340	−0.362	0.715	−0.711	1.000	−0.663	0.451
CO2	0.036	0.642	−0.573	0.699	−0.663	1.000	−0.437
NMIG	−0.077	−0.591	0.254	−0.280	0.451	−0.437	1.000

and

Table A2. Correlation matrix for Med8.

	LE	GDP	RENC	FF	FDI	CO2	NMIG
LE	1.000	0.761	0.174	−0.139	0.166	−0.074	0.005
GDP	0.761	1.000	0.176	−0.278	0.040	0.218	−0.107
RENC	0.174	0.176	1.000	−0.428	−0.230	−0.176	−0.053
FF	−0.139	−0.278	−0.428	1.000	0.218	0.163	0.450
FDI	0.166	0.040	−0.230	0.218	1.000	−0.276	0.278
CO2	−0.074	0.218	−0.176	0.163	−0.276	1.000	−0.177
NMIG	0.005	−0.107	−0.053	0.450	0.278	−0.177	1.000

). All pairwise correlations are below 0.90, indicating no critical collinearity issues.

4. Results and Discussions

The Preston curve is an empirical relationship that demonstrates a positive association between LE and income per capita across countries. It was introduced by Preston in 1975 and is a key concept in demography and economics, highlighting the link between economic conditions and health outcomes [63]. The curve shows that, in general, countries with higher income per capita tend to have higher LE. This relationship is particularly strong for low- and middle-income countries. The Preston curve exhibits diminishing returns at higher levels of income. While LE increases rapidly with rising income at lower levels, the rate of increase slows down as income rises. In other words, additional gains in income have smaller impacts on LE in wealthier countries. The curve may shift over time due to improvements in public health, medical technology, education, and other factors that influence LE independently of income.

Throughout this paper, we work with the variables expressed in their natural logarithm form. The Preston curve equation is

$$\mathrm{L}\mathrm{E}=\mathrm{f}\left(\mathrm{G}\mathrm{D}\mathrm{P}\right)$$

(7)

Our research extends the Preston curve model by incorporating other predictors, as mentioned earlier. Consequently, the following model is employed:

$$\mathrm{L}\mathrm{E}=\mathrm{f}(\mathrm{G}\mathrm{D}\mathrm{P},\mathrm{R}\mathrm{E}\mathrm{N}\mathrm{C},\mathrm{F}\mathrm{F},\mathrm{F}\mathrm{D}\mathrm{I},\mathrm{C}\mathrm{O}2,\mathrm{N}\mathrm{M}\mathrm{I}\mathrm{G})$$

(8)

4.1. First- and Second-Generation PURT and Panel ARDL

The first- and second-generation PURTs in

Table 2. First-generation PURT—Med8.

At Levels
	LE	RENC	FF	GDP	FDI
Unit root (Common Unit Root Process)
LLC	−7.09 *** (0.000)	0.60 (0.727)	0.24 (0.596)	−1.74 ** (0.04)	−2.23 ** (0.012)
Unit root (Individual Unit Root Process)
IPS	−3.13 *** (0.198)	2.58 (0.995)	2.90 (0.998)	0.403 (0.656)	−3.08 *** (0.001)
ADF	37.64 *** (0.001)	6.02 * (0.987)	3.63 (0.999)	9.40 (0.656)	34.60 *** (0.004)
At first difference
Unit root (Common Unit Root Process)
LLC	−5.08 *** (0.000)	−5.77 *** (0.000)	−5.43 *** (0.000)	−6.06 *** (0.000)	−9.87 *** (0.000)
Unit root (Individual Unit Root Process)
IPS	−8.79 *** (0.000)	−7.76 *** (0.000)	−8.03 *** (0.000)	−7.63 *** (0.000)	−12.61 *** (0.000)
ADF	103.74 *** (0.000)	88.85 *** (0.000)	91.90 *** (0.000)	87.73 *** (0.000)	153.91 *** (0.000)
At levels
	CO2	NMIG
LLC	2.25 (0.987)	−5.67 *** (0.000)
Unit root (Individual Unit Root Process)
IPS	3.23 (0.999)	−4.99 *** (0.000)
ADF	4.21 (0.999)	58.79 *** (0.000)
At first difference
Unit root (Common Unit Root Process)
LLC	−6.11 *** (0.000)	−10.42 *** (0.000)
Unit root (Individual Unit Root Process)
IPS	−7.00 *** (0.000)	−10,52 *** (0.000)
ADF	78.47 *** (0.000)	58.79 *** (0.000)

*, **, *** significant at 10%, 5% and 1% level.

,

Table 3. First-generation PURT—Med4.

At Levels
	LE	RENC	FF	GDP	FDI
Unit root (Common Unit Root Process)
LLC	−2.47 *** (0.006)	−1.49 * (0.068)	−1.57 * (0.057)	−2.10 ** (0.022)	−1.38 * (0.083)
Unit root (Individual Unit Root Process)
IPS	−0.84 (0.198)	−1.66 ** (0.047)	−1.17 (0.12)	1.40 (0.920)	−1.58 * (0.056)
ADF	11.28 (0.186)	14.51 * (0.069)	13.97 * (0.082)	8.04 (0.429)	14.45 * (0.07)
At first difference
Unit root (Common Unit Root Process)
LLC	−13.23 *** (0.000)	−8.59 *** (0.000)	−6.04 *** (0.000)	−3.44 *** (0.000)	−6.34 *** (0.000)
Unit root (Individual Unit Root Process)
IPS	−9.87 *** (0.000)	−9.60 *** (0.000)	−7.10 *** (0.000)	−6.09 *** (0.000)	−7.06 *** (0.000)
ADF	73.74 *** (0.000)	83.44 *** (0.000)	59.42 *** (0.000)	50.84 *** (0.000)	58.96 *** (0.000)
At levels
	CO2	NMIG
LLC	−0.75 (0.224)	0.717 (0.763)
Unit root (Individual Unit Root Process)
IPS	0.44 (0.679)	−0.369 (0.355)
ADF	4.02 (0.855)	8.498 (0.386)
At first difference
Unit root (Common Unit Root Process)
LLC	−6.63 *** (0.000)	−6.59 *** (0.000)
Unit root (Individual Unit Root Process)
IPS	−7.28 *** (0.000)	−4.08 *** (0.000)
ADF	61.24 *** (0.000)	31.62 *** (0.000)

*, **, *** significant at 10%, 5% and 1% level.

,

Table 4. Second-generation PURT CIPS—Med8.

At Levels
	LE	RENC	FF	GDP	FDI
CIPS	−2.54 (<0.05)	−2.35 (<0.05)	−2.10 (≥0.10)	−1.42 (≥0.10)	−2.94 (<0.01)
At first difference
CIPS	−4.45 (<0.01)	−5.64 (<0.01)	−5.57 (<0.01)	−4.61 (<0.01)	−4.46 (<0.01)
At levels
	CO2	NMIG
CIPS	−1.69 (>0.10)	−1.8 (≥0.10)
CIPS	−4.38 (<0.01)	−6.59 (<0.01)

and

Table 5. Second-generation PURT CIPS—Med4.

At Levels
	LE	RENC	FF	GDP	FDI
CIPS	−1.09 (≥0.10)	−0.41 (≥0.10)	−0.08 (≥0.10)	−1.42 (≥0.10)	−1.10 (≥0.10)
At first difference
CIPS	−4.45 (<0.01)	−2.37 (<0.05)	−6.61 (<0.01)	−3.56 (<0.05)	−3.89 (<0.01)
At levels
	CO2	NMIG
CIPS	−1.42 (>0.10)	−2.27 (<0.10)
CIPS	−4.71 (<0.01)	−2.98 (<0.01)

show that all variables become stationary after first differencing. Hence, we apply the panel ARDL models as in

Table 6. Panel ARDL(3,3,3,3,3,3,3) for Med8.

Variable	Coefficient	Std. Error	t-Statistic	Prob. *
	Long-Run Equation
RENC	0.046981	0.012824	3.663477	0.000 ***
FF	0.103384	0.060767	1.701304	0.092 *
GDP	0.104497	0.01803	5.795577	0
NMIG	−0.003969	0.002007	−1.977823	0.051 *
FDI	−0.004147	0.001678	−2.471631	0.015 **
CO2	0.045902	0.018002	2.549821	0.012 **
	Short-Run Equation
COINTEQ01	−0.11979	0.052347	−2.288403	0.024 **
D(LE(−1))	−0.253996	0.122385	−2.07539	0.040 **
D(LE(−2))	−0.173416	0.07738	−2.24111	0.027 **
D(RENC)	−0.006997	0.00513	−1.363948	0.176
D(RENC(−1))	0.002341	0.008601	0.272156	0.786
D(RENC(−2))	−0.017265	0.00468	−3.689231	0.000 ***
D(FF)	0.068425	0.132788	0.515294	0.607
D(FF(−1))	−0.202505	0.109211	−1.854246	0.067 *
D(FF(−2))	−0.061023	0.055301	−1.103473	0.272
D(GDP)	0.015147	0.025281	0.59916	0.550
D(GDP(−1))	−0.019163	0.017164	−1.116459	0.267
D(GDP(−2))	−0.027077	0.016699	−1.621473	0.108
D(NMIG)	0.001595	0.00112	1.424609	0.157
D(NMIG(−1))	−0.000132	0.000959	−0.138074	0.890
D(NMIG(−2))	0.003013	0.001733	1.738516	0.085
D(FDI)	0.000658	0.000563	1.16947	0.245
D(FDI(−1))	−0.000444	0.00055	−0.805775	0.422
D(FDI(−2))	−0.000277	0.000523	−0.529247	0.597
D(CO2)	−0.014736	0.014356	−1.026452	0.307
D(CO2(−1))	0.026362	0.014508	1.817157	0.072 *
D(CO2(−2))	0.000958	0.00627	0.152712	0.879
C	0.325193	0.14087	2.308454	0.023 **

*, **, *** significant at 10%, 5%, and 1% level.

and

Table 7. Panel ARDL(3,1,1,1,1,1,1) for Med4.

Variable	Coefficient	Std. Error	t-Statistic	Prob. *
	Long-Run Equation
RENC	−0.168975	0.015889	−10.63451	0 ***
FF	−0.265278	0.050585	−5.244259	0 ***
GDP	−0.022717	0.017659	−1.286428	0.201
NMIG	0.003423	0.001515	2.259659	0.026 **
FDI	−0.008827	0.003443	−2.563861	0.012 **
CO2	0.07975	0.025336	3.14766	0.002 ***
	Short-Run Equation
COINTEQ01	−0.243438	0.131278	−1.854371	0.067 *
D(LE(−1))	−0.07137	0.080314	−0.888636	0.376
D(LE(−2))	0.028723	0.129095	0.222496	0.824
D(RENC)	0.028552	0.020284	1.407596	0.162
D(FF)	0.111252	0.100718	1.104583	0.272
D(GDP)	0.124177	0.06434	1.930012	0.056 *
D(NMIG)	−0.000458	0.001141	−0.401704	0.688
D(FDI)	−0.000593	0.002571	−0.230548	0.818
D(CO2)	−0.048852	0.027922	−1.749582	0.083
C	1.486862	0.798437	1.862216	0.065 *

*, **, *** significant at 10%, 5%, and 1% level.

, and Figure 2 and Figure 3.

Figure 2 presents the AIC values used to determine the optimal lag length for the panel ARDL model in the Med8 group. The minimum AIC value corresponds to the model with lag order (3,3,3,3,3,3,3), which balances goodness-of-fit with model complexity. This lag structure ensures that both short-run and long-run dynamics are adequately captured while minimizing information loss. The selected model specification is then used in Table 6 for the panel ARDL estimation.

Before presenting the panel ARDL estimates for Med8 and Med4 groups in Table 6 and Table 7, we outline some considerations for interpreting the results. The focus is on statistically significant long-run coefficients, as these reflect persistent effects of the explanatory variables on LE. Together with each coefficient, we report robust standard errors, which quantify the degree of sampling variability. Smaller standard errors indicate more precise estimates, while larger ones signal greater uncertainty around the coefficient. In our results, the standard errors are generally within acceptable ranges and consistent with expectations for macro-panel data, reinforcing the reliability of significant coefficients. Significance is denoted at conventional levels (1%, 5%, 10%). To address potential multicollinearity, we examined the correlation matrices (Appendix A, Table A1 and Table A2). The absence of extreme correlations (above 0.9) suggests that multicollinearity does not substantially bias the coefficient estimates.

As Table 6 reports, in the long term, a 1% increase in RENC leads to a 0.047% increase in LE. This positive relationship suggests that increasing RENC improves public health by reducing pollution and promoting environmental sustainability. A 1% increase in FF causes a 0.103% increase in LE in the long term. This positive long-term effect may reflect the role of FF in driving economic growth and improvements in living standards, which indirectly benefit LE. However, it could also be a sign that FF-dependent economies, while improving LE through infrastructure and industrialization, might not have yet fully experienced the negative health effects of environmental degradation. A 1% increase in GDP corresponds to a 0.104% rise in LE in the long run. This is consistent with the theory that economic growth improves living conditions, healthcare access, and education, contributing to longer LE. A 1% increase in NMIG causes a 0.004% decrease in LE. This slight negative effect could stem from the strain that increased migration places on public services and infrastructure, particularly if the inflow of migrants is not accompanied by sufficient investments in healthcare and social services. A 1% increase in FDI is associated with a 0.004% reduction in LE. This suggests that FDI, in carbon-intensive or resource-extractive industries, might harm the environment and public health over time, thereby reducing LE. A 1% increase in CO₂ emissions corresponds to a 0.046% increase in LE. This positive relationship might reflect the benefits of economic growth and industrialization, which accompany higher emissions in the short to medium term, despite their eventual environmental costs.

The ECT is negative and significant, belonging to [−1,0]. This indicates a speed of adjustment of about 11.98% towards the long-run equilibrium each period. This means that short-run deviations from the long-term LE trend will gradually adjust back to equilibrium at a moderate pace. The negative values of D(LE(−1)), D(LE(−2)) show that past decreases in LE lead to further reductions in the short term, reflecting a mean-reverting process where temporary shocks persist for a while before correcting. The short-term effects of RENC are mixed, with the first lag showing an insignificant positive effect and the second lag showing a significant negative effect. This could indicate that the immediate adoption of RES might come with transitional challenges or costs, but in the long term, the benefits become clearer.

In the short term, FF has a positive effect in the first period but turns negative in subsequent lags. This suggests that while fossil fuels may initially support economic activity and LE, their environmental and health costs quickly start to outweigh those benefits, causing a negative impact on LE. In the short run, changes in GDP have an insignificant effect on LE. The initial positive effect is followed by negative coefficients in subsequent periods, indicating that the benefits of economic growth take time to manifest in health improvements. NMIG has a small and mostly insignificant short-term impact on LE. However, the third lag shows a positive effect, suggesting that while the immediate effects of migration might not be substantial, longer-term integration of migrants could benefit the overall population. FDI has an insignificant short-term impact on LE, with fluctuating signs. This implies that FDI, while economically beneficial, may not translate directly into immediate improvements in health and LE. The short-run effects of CO₂ emissions on LE are mixed, with a negative initial impact followed by a positive one in the second lag. This suggests that the immediate environmental and health costs of higher emissions may be offset in subsequent periods by the economic growth and development they drive.

Figure 3 shows the AIC-based model selection process for the Med4 group. The optimal lag order selected by the lowest AIC is (3,1,1,1,1,1,1). This more parsimonious structure compared with Med8 reflects differences in data dynamics and sample variability between the two regional groups. The selected lag structure was used for estimating the panel ARDL model presented in Table 7, ensuring efficiency without overfitting.

Alternative lag structures have been tested during the ARDL estimation process. Due to the time length and degrees of freedom constraints, it was not possible to estimate models with lags higher than 3. Lags above 3 led to overparameterization or non-convergence, especially for the Med4 group with a shorter effective sample. Therefore, the choice of lag length was driven by both statistical diagnostics (e.g., AIC) and practical feasibility.

As seen in Table 7, in the long term, a 1% increase in RENC is associated with a 0.169% decrease in LE. This unexpected negative relationship could suggest challenges in transitioning to RES in Med4 countries. It also reflects inefficiencies in RES infrastructure, or perhaps economic and social disruptions associated with reducing dependence on FF. A 1% increase in FF results in a 0.265% decrease in LE in the long run. This negative impact reflects the harmful effects of pollution and environmental degradation caused by FF, which diminish air quality, increase respiratory illnesses, and lower LE. A 1% increase in NMIG leads to a 0.0034% increase in LE. This positive impact indicates that NMIG in these countries brings potential benefits to LE. A 1% increase in FDI reduces LE by 0.009% in the long run. This negative impact results from labor exploitation, environmental degradation, or displacement caused by FDI. A 1% increase in CO₂ emissions results in a 0.080% increase in LE in the long term. While counterintuitive, this could be due to industrial development and infrastructure improvements that raise living standards, though at an environmental cost. The ECT is negative and significant, indicating a return to long-term equilibrium after short-term deviations. Approximately 24.34% of the short-term disequilibrium is corrected each period. A 1% increase in RENC leads to a 0.029% increase in LE in the short term, suggesting a more immediate positive impact of RES on health compared with the long-term effects. A 1% increase in FF raises LE by 0.111% in the short term, reflecting the potential benefits of energy availability for economic activity and healthcare services despite its long-term harmful effects. A 1% increase in GDP boosts LE by 0.124% in the short term, suggesting that economic growth quickly translates into better health outcomes, likely through improved infrastructure, healthcare access, and living standards. In the short term, a 1% increase in CO₂ emissions reduces LE by 0.049%. This result aligns with the detrimental immediate effects of pollution on health.

4.2. Comparative Discussion for Med8 and Med4 Countries

Med8 countries, RENC has a positive impact on LE. It follows that the shift toward REs like solar and wind is contributing to better health outcomes. This could be due to reduced pollution and cleaner air, leading to improved public health. Interestingly, in the Med4 countries, RENC has a negative impact on LE. This unexpected outcome may be linked to economic or infrastructural challenges in these countries, where the transition to RES might be more disruptive or costly, potentially affecting public services, health infrastructure, or employment negatively.

The relationship between FF and LE in Med8 countries is positive, although not as strongly significant. This might suggest that, while FF are environmentally harmful, they still contribute to economic activities that support better healthcare systems and infrastructure. In contrast, FF in the Med4 countries shows a negative impact on LE. This is more in line with what one might expect, as reliance on FF leads to greater pollution, which directly harms public health and reduces LE. These countries may also face more severe pollution-related health problems compared with the Med8 group. Also, economic growth has a positive influence on LE in the Med8 countries. This suggests that as these economies expand, healthcare and education services are improved. This enhances the quality of life and longevity. In contrast, economic growth does not seem to have the same positive influence on the Med4 group. In fact, the relationship is weakly negative. This could reflect that economic growth in these countries might not be evenly distributed or invested in public health and infrastructure, meaning the benefits of growth are not being felt broadly by the population. NMIG shows a negative effect on LE in the Med8 group, suggesting that high levels of immigration might strain public resources or healthcare systems in the short term, reducing overall LE. In the Med4 group, NMIG has a positive effect, indicating that the influx of migrants may have a more beneficial role.

The long-run positive association between CO₂ emissions and LE in both Med8 and Med4 country groups seems paradoxical. However, similar patterns have been noted in recent studies. For instance, ref. [64] finds that while CO₂ emissions generally reduce LE in emerging economies, certain developing countries experience a positive association, likely due to CO₂ emissions reflecting imported consumption rather than domestic production. Likewise, ref. [65] identifies a non-linear relationship between CO₂ and LE in India, where moderate emissions associated with economic growth and industrialization initially support better health infrastructure and higher incomes, indirectly increasing LE, up to a threshold beyond which the adverse effects dominate. These findings indicate that in some economies, higher CO₂ emissions may signal developmental progress that temporarily enhances health outcomes despite growing environmental costs.

The divergence between Med8 and Med4 suggests that while the Preston curve holds for more developed Mediterranean countries (Med8), less developed or transitioning economies (Med4) do not follow the same pattern. In Med4, factors beyond income are playing a more dominant role in determining LE, emphasizing the need for broader improvements in healthcare systems, public health initiatives, and social development.

As seen in

Table 8. Granger causality test for Med8.

NH:	F-Statistic	Prob.	Conclusion
RENC nGc LE	1.20	0.3
LE nGc RENC	6.18	0.002 ***	LE→RENC
FF nGc LE	3.13	0.045 **	FF→LE
LE nGc FF	2.67	0.071 *	LE→FF
GDP nGc LE	4.88	0.008 ***	GDP→LE
LE nGc GDP	3.12	0.045 **	LE→GDP
NMIG nGc LE	0.05	0.947
LE nGc NMIG	0.26	0.770
FDI nGc LE	1.02	0.360
LE nGc FDI	1.52	0.220
CO2 nGc LE	2.99	0.051 *	CO2→LE
LE nGc CO2	13.53	3.00 × 10−6 ***	LE→CO2
FF nGc RENC	0.16	0.845
RENC nGc FF	0.18	0.833
GDP nGc RENC	1.99	0.138
RENC nGc GDP	0.96	0.380
NMIG nGc RENC	0.44	0.643
RENC nGc NMIG	0.32	0.722
FDI nGc RENC	1.40	0.246
RENC nGc FDI	1.92	0.148
CO2 nGc RENC	0.03	0.966
RENC nGc CO2	0.78	0.4562
GDP nGc FF	2.77	0.064 *	GDP→FF
FF nGc GDP	1.29	0.277
NMIG nGc FF	0.56	0.571
FF nGc NMIG	10.59	4.00 × 10−5 ***	FF→NMIG
FDI nGc FF	1.45	0.235
FF nGc FDI	0.71	0.487
CO2 nGc FF	3.46	0.032
FF nGc CO2	0.10	0.904
NMIG nGc GDP	0.01	0.987
GDP nGc NMIG	1.10	0.333
FDI nGc GDP	0.18	0.832
GDP nGc FDI	0.22	0.801
CO2 nGc GDP	1.66	0.191
GDP nGc CO2	3.78	0.024 **	GDP→CO2
FDI nGc NMIG	2.75	0.065
NMIG nGc FDI	1.21	0.292
CO2 nGc NMIG	2.00	0.137
NMIG nGc CO2	0.56	0.567
CO2 nGc FDI	1.18	0.306
FDI nGc CO2	0.84	0.429

nGc stands for “does not homogeneously cause”; *, **, *** significant at 10%, 5%, and 1% level.

, higher LE leads to greater RENC. Longer-living populations are likely to push for better environmental policies and cleaner energy sources. FF influences LE in the Med8 countries, likely in a negative way. Economic growth has a strong and positive influence on LE. As GDP rises, resources for healthcare, education, and public services increase, contributing to longer and healthier lives. Higher LE positively influences economic growth. A healthier, longer-living workforce contributes to greater productivity, innovation, and investment potential, which in turn drives GDP growth. CO₂ emissions negatively affect LE, due to pollution and environmental degradation associated with high emissions. Higher LE leads to reduced CO₂ emissions. GDP causes an increase in CO₂ emissions because growing economies consume more energy and industrial activity, which results in higher emissions unless offset by sustainable practices. In the Med8, rising GDP comes with environmental costs, including greater CO₂ output. Higher FF influences NMIG, possibly due to the environmental degradation and economic opportunities associated with FF-driven industries.

As seen in

Table 9. Granger causality test for Med4.

NH:	F-Statistic	Prob.	Conclusion
RENC nGc LE	3.85	0.051 *	RENC→LE
LE nGc RENC	14.26	0 ***	LE→RENC
FF nGc LE	1.18	0.278
LE nGc FF	6.84	0.009 ***	LE→FF
GDP nGc LE	3.00	0.085 *	GDP→LE
LE nGc GDP	0.45	0.503
NMIG nGc LE	1.35	0.246
LE nGc NMIG	0.01	0.907
FDI nGc LE	1.71	0.192
LE nGc FDI	0.19	0.663
CO2 nGc LE	0.17	0.672
LE nGc CO2	22.51	5.00 × 10−6 ***	LE→CO2
FF nGc RENC	2.88	0.091 *	FF→RENC
RENC nGc FF	0.28	0.596
GDP nGc RENC	12.59	0 ***	GDP→RENC
RENC nGc GDP	0.06	0.79
NMIG nGc RENC	7.69	0.006 ***	NMIG→RENC
RENC nGc NMIG	0.01	0.888
FDI nGc RENC	7.30	0.007 ***	FDI→RENC
RENC nGc FDI	3.14	0.078 *	RENC→FDI
CO2 nGc RENC	0.11	0.735
RENC nGc CO2	0.24	0.618
GDP nGc FF	3.90	0.050 **	GDP→FF
FF nGc GDP	0.14	0.702
NMIG nGc FF	1.85	0.175
FF nGc NMIG	0.98	0.322
FDI nGc FF	7.94	0.005 ***	FDI→FF
FF nGc FDI	2.47	0.118
CO2 nGc FF	0.1177	0.7321
FF nGc CO2	0.01	0.908
NMIG nGc GDP	1.86	0.174
GDP nGc NMIG	11.41	0.001 ***	GDP→NMIG
FDI nGc GDP	2.80	0.096 *	FDI→GDP
GDP nGc FDI	0.01	0.909
CO2 nGc GDP	2.13	0.146
GDP nGc CO2	11.55	0 ***	GDP→CO2
FDI nGc NMIG	2.17	0.142
NMIG nGc FDI	0.77	0.380
CO2 nGc NMIG	3.01	0.084 *	CO2→NMIG
NMIG nGc CO2	1.17	0.280
CO2 nGc FDI	0.84	0.359
FDI nGc CO2	2.38	0.125

“nGc“ stands for “does not homogeneously cause”; *, **, *** significant at 10%, 5%, and 1% level.

, higher RENC positively impacts LE in the Med4 countries. Higher LE also leads to increased RENC. FF drives RENC in these countries, possibly reflecting the need for energy diversification as the environmental and economic costs of FF rise. Economic growth leads to increased RENC, likely due to greater financial resources available for investment in clean energy infrastructure and technologies in the Med4 region. NMIG influences RENC, possibly through increasing energy demand as populations grow or through policy changes aimed at supporting sustainable development in response to social challenges posed by migration. FDI supports RENC, likely because FDI brings in capital for large-scale clean energy projects, especially in developing regions like the Med4. RENC attracts FDI, indicating that countries focusing on clean energy development are becoming attractive destinations for international investors. Economic growth positively impacts LE, as higher GDP allows for better healthcare. Higher LE may lead to reduced FF, as longer-living populations become more environmentally conscious and demand cleaner energy sources. Economic growth drives FF, reflecting that growing economies in the Med4 still rely heavily on FF, even as RES becomes more important. FDI supports FF, potentially indicating that foreign investors continue to invest in traditional energy sectors in some Med4 countries, despite growing interest in RES. Economic growth increases CO₂ emissions, which is consistent with the idea that industrial expansion and energy consumption, driven by FF, contribute to environmental degradation in these countries. Higher LE reduces CO₂ emissions, suggesting that populations with longer lifespans demand stricter environmental policies and better air quality. CO₂ emissions drive NMIG, possibly due to environmental degradation and climate-related factors that force people to move. Economic growth influences NMIG, as prosperous economies attract migrants seeking employment and better economic opportunities. FDI positively influences economic growth in the Med4, suggesting that FDI is an important driver of development and economic expansion.

To answer the research questions (RQs) formulated in the Section 1, we include

Table 10. Main findings, directions, and interpretation.

Variable	Med8 LE (Long Run)	Med4 LE (Long Run)	Direction and Interpretation
RENC	+0.047% (***)	–0.169% (***)	Helps health in Med8, harms in Med4 due to transition challenges
FF	+0.103% (*)	–0.265% (***)	Still supports infrastructure in Med8, but damages health in Med4
GDP	+0.104% (***)	–0.023% (ns)	Growth improves LE in Med8, not in Med4
NMIG	–0.004% (*)	+0.0034% (**)	Strain in Med8; benefit in Med4
FDI	–0.004% (**)	–0.009% (**)	Harmful in both due to dirty investments
CO2	+0.046% (**)	+0.08% (***)	Proxy for development, with long-term trade-offs

*, **, *** significant at 10%, 5%, and 1% level; “ns” stands for “not significant”

, which synthesizes the main results.

5. Conclusions and Policy Implications

Our research analyzed the determinants of LE in Mediterranean countries using a panel ARDL approach applied separately to Med8 and Med4 groups. Several results align with prior expectations (e.g., GDP positively affecting LE in Med8), while others reveal counterintuitive patterns requiring further reflection.

The long-run positive relationship between CO₂ and LE, as well as FF and LE in certain models, appears paradoxical. These outcomes may reflect the role of CO₂ and FF as proxies for industrial development, infrastructure expansion, and improved public services that can enhance LE in the short to medium term. In such contexts, increased GDP and energy availability often lead to improvements in healthcare access, water systems, and transport. However, these benefits may be temporary and eventually offset by environmental degradation, suggesting a nonlinear or inverted U-shaped dynamic similar to the EKC framework.

The negative effect of RENC on LE observed in Med4 may reflect transition-related inefficiencies, weak infrastructure, or higher energy costs affecting household well-being. In contrast, RENC has a positive and significant effect in Med8, where clean energy transitions are more mature and institutionally supported.

The roles of NMIG and FDI are heterogeneous. NMIG contributes positively to LE in Med4, potentially through labor contributions and demographic rejuvenation, while negatively affecting LE in Med8, where integration pressures may burden public services. FDI negatively affects LE in both groups, possibly due to environmentally or socially adverse investments in countries with limited regulatory frameworks. These findings align with previous research emphasizing the economic and demographic relevance of migration in sustainable development processes, as demonstrated by ref. [66].

Policy measures must therefore be tailored to regional dynamics. In Med8, where GDP and RENC improve LE, policies should support energy transition pathways while aligning them with public health strategies. Environmental oversight in FDI and inclusive healthcare for migrant populations are also necessary. In Med4, energy policy should focus on removing implementation bottlenecks and increasing the efficiency of RENC deployment. Stronger institutional regulation is also needed to align FDI and migration with health and sustainability objectives.

Our current research is limited by its exclusion of institutional variables such as healthcare system quality, political stability, or inequality, which may mediate the identified effects. These omissions constrain the external validity of the results. Future research should integrate such variables and apply robustness checks (e.g., nonlinear ARDL or Generalized Method of Moments (GMM)) to enhance empirical credibility and policy relevance.