Assessing the Impact of Smart and Green Transition Policies on Spatial and National Income Inequalities in EU Countries

Abstract

Green and digital transitions represent a dual strategic objective for the European Union (EU), requiring behavioral changes from citizens, markets, and state institutions. To support these transformations, the EU has developed an extensive policy framework that is backed by significant financial instruments. However, the existing research suggests that these transitions may exacerbate both spatial and socioeconomic inequalities, depending on country-specific conditions and institutional capacities. This paper investigates how environmental and technological contexts, alongside EU-transition-related policies, influence regional and income inequalities within the selected EU countries. Using panel data covering the period 2007–2020 and employing a Generalized Least Squares (GLS) estimator, the present study reveals the complex relationship between structural conditions, policy designs, and inequality outcomes. The results show that smart and green policies tend to mitigate spatial inequalities, though they are found to be less effective in addressing income inequalities. By contrast, the contextual dynamics of the twin transition, such as skill intensity, digital adoption, and emissions, exhibit mixed effects, sometimes reinforcing inequality. The findings underline the urgency of designing inclusive EU policies that combine green and smart transition measures while accounting for country- and region-specific challenges. Such an integrated approach is essential for ensuring that the twin transition strengthens social cohesion in Europe, rather than undermining it.

1. Introduction

In recent decades, rapid environmental and technological changes have significantly reshaped societies. Climate change poses a major global threat, marked by rising temperatures, shifting rainfall patterns, and increased extreme weather events, leading to financial losses, biodiversity loss, and growing inequality [1,2]. Concurrently, technological progress has accelerated, offering opportunities but also deepening the challenges related to inclusion, adaptability, and equal access [3,4]. Dealing with those challenges requires a green and digital transition—often called the twin transition. The implementation of these two drivers involves significant challenges being faced in ensuring social and spatial cohesion [5].

Climate change and natural disasters exacerbate inequality for three main reasons: poorer countries and populations are more exposed, suffer proportionally greater losses, and have fewer resources for recovery [6]. Wealthier individuals can better adapt to climate risks such as extreme heat, droughts, wildfires, and flooding [7]. Global warming has widened income gaps, with temperature showing a parabolic relationship with economic growth [8]. Using panel data, studies have shown short-term increases in inequality following disasters, though long-term effects may fade [9,10]. At the same time, regional inequality may also be noticed, especially as regions in Southern and Southeastern Europe show lower average socioeconomic status and have higher proportions of elderly people [11]. Negative correlations between socioeconomic status and local environmental conditions have been noticed as well [12].

Paglialunga et al. [6] find that temperature and precipitation anomalies worsen inequalities; Jalles [13] linked the rising climate risks to global income inequality. Gilli et al. [14] argue that climate damages and mitigation costs disproportionately affect low-income countries, while Taconet et al. [15] caution that the extent of inequality impacts depends on multiple contextual factors such as the level of development and the reliance on the agricultural sector. In summary, a systematic review by Mejean et al. [16] confirmed that most studies report the negative effect of climate change on inequality.

On the other hand, the digitalization of our economies has been viewed as a tool to reduce global inequality; however, it can also be noticed that digital technologies are leading to a higher concentration of power and wealth [17,18]. Therefore, digitalization’s impact on income inequality is debated. In spite of that, it can reduce inequality by improving productivity, creating new job opportunities, and expanding access to resources and markets. On the other hand, it may widen inequality by favoring skilled over unskilled workers, replacing routine jobs through automation, and deepening digital divides across income groups and regions [19]. Before the digital revolution, inequality, as measured by the Gini coefficient, decreased and then showed a great increase upon digitalization [20]. In the systematic review by Bauer [21], it is stated that Information and Communication Technologies (ICTs) interact with economic, technological, and political factors in ways that can either widen or reduce income inequality, depending on which forces dominate and the surrounding political and institutional context.

In addition, Afzal et al. [22] found that technological penetration’s effect on income inequalities is conditioned by income levels, with inequalities increasing when technology penetrates further in low-income countries. In the same vein, Wang and Shen [23], in their study on 97 countries, found a positive effect of digitalization on income equality. Nevertheless, they showed that this effect was not stable across different income groups of countries. In addition, Consoli et al. [24], studying the intra-regional inequalities in EU countries, found that the impact of digital skills on inequality differs notably by income group: digitalization tends to widen inequalities among lower-income populations, while it helps reduce them among higher-income groups. Therefore, changes in technology are not delivering their full potential in terms of increasing productivity and economic growth; instead, they can drive inequality up and lead to a more unequal distribution of capital and labor income [25].

The above-mentioned dual shifts—green and digital—form the basis of the EU’s “twin transition”, which aims to achieve climate neutrality and global competitiveness. Therefore, the EU has developed an extensive policy framework, in addition to assigning relevant supporting funds to that end [26]. To further ensure that these dynamics will not harm social and regional cohesion, the EU has integrated those targets in its Cohesion Policy (CP), and is directing substantial funds toward those people and places which are facing the greatest challenges [27].

While the existing literature has examined either the dynamics or the policies of the twin transition, few studies integrate both dimensions within a single analytical framework. This paper addresses this gap by evaluating how green and digital transitions affect inequality in selected EU countries, and how Cohesion Policy interventions supporting these transitions influence social cohesion. The analysis considers both spatial and income inequalities, thus providing a comprehensive overview of how the twin transition has shaped cohesion across the EU during the two programming periods up to 2020. Importantly, the study explicitly recognizes that green and digital transitions are not isolated processes; rather, they are interdependent dimensions of a broader socioeconomic transformation. Rather than treating environmental, technological, and policy factors separately, this study models their combined and interacting effects on spatial and income inequalities.

Ultimately, the analysis seeks to answer the following research questions:

Does the evolution of smart and green transitions affect spatial and socioeconomic inequalities? If so, how?

What is the impact of policies supporting smart and green transitions on spatial and socioeconomic inequalities?

The remainder of the paper is structured as follows: Section 2 presents the literature review and explains the research hypotheses in greater detail. Section 3 outlines the methodological framework and data sources used to address the research questions. Section 4 presents the main findings, and the Section 5 assesses them in relation to the literature. Finally, the Section 6 summarizes how we can ensure a fairer twin transition, and makes policy recommendations.

2. Literature Review

The goal of smart transition is to make Europe digitally sovereign, inclusive, and globally competitive. The EU promotes smart transition—centered on digitalization, innovation, and technological modernization—through a comprehensive policy framework that combines funding, strategic planning, and regulation. The contemporary flagship program driving the digital transition is the Digital Decade Policy Programme 2030. This sets clear targets for digital skills, infrastructure, business digitalization, and e-government. Major funding instruments such as the Digital Europe Programme, Horizon Europe, and the Recovery and Resilience Facility (RRF) provide substantial resources for digital technologies, research, and infrastructure [28,29].

On the green transition, the EU aims to reach climate neutrality by 2050, with a transitional goal of reducing greenhouse gas (GHG) emissions by 55% by 2030. The European Green Deal and Global Gateway initiatives are directing substantial investments to Member States in the environmental domain [30,31].

This twin transition can deepen disparities across regions and social groups. To counter this risk, the EU has embedded the twin transition within the CP, which promotes social and territorial cohesion. Significant funding is directed to less developed and more vulnerable regions to help them adapt. Smart Specialization Strategies ensure that investments align with local strengths. Key funds like the European Regional Development Fund (ERDF) and the Just Transition Fund support clean energy, digital infrastructure, and upskilling initiatives. The CP also emphasizes digital inclusion, support for workers in carbon-intensive sectors, and broader goals such as gender equality and accessibility [27,32].

Table 1. Targets and respective budget of Cohesion Policy supporting smart and green transitions for 2014–2020 and 2021–2027 programming periods.

Transition Dimension	Programming Period
	2014–2020		2021–2027
	Thematic Objectives	Budget in bn. EUR (%)	Specific Objectives	Budget in bn. EUR (%)
Smart	TO1: Strengthening research, technological development and innovation	97 (24%)	PO1: A smarter Europe	75 (20%)
	TO2: Enhancing access to, and use and quality of ICT
	TO3: Enhancing the competitiveness of small- and medium-sized enterprises
Green	TO4: Supporting the shift towards a low-carbon economy in all sectors	69 (17%)	PO2: A greener, low-carbon Europe	96 (25%)
	TO5: Climate change adaptation and risk management
	TO6: Protecting the environment and promoting resource efficiency

Sources: [27,35]; own elaboration.

outlines how smart and green transition targets are integrated into CP objectives for 2014–2020 and 2021–2027. It also estimates funding allocations and identifies main financial instruments supporting an inclusive transition. For 2021–2027, two of the five policy objectives directly address smart and green goals. In 2014–2020, the categorization is less direct. Following the Ninth Cohesion Report [29], we align TOs 1–3 with smart priorities and TOs 4–6 with green priorities. The data show smart-related funding dominated in 2014–2020, but this trend reverses in 2021–2027 as green priorities receive greater investment. The main funding instruments include the European Regional Development Fund (ERDF), the Cohesion Fund, and European Social Fund Plus (ESF+), complemented by the Just Transition Fund for regions shifting away from fossil fuels. Those funding mechanisms are part of the European Structural and Investment Fund (ESIF) [33,34].

Cohesion considerations related to the twin transitions—green and smart—have long existed in the EU policy framework. However, the effectiveness of these policies in addressing inequalities remains debated. Recent research shows that climate change mitigation is closely linked to existing and emerging patterns of inequality [36]. The main impacts of climate mitigation policies and increasing energy prices are primarily noticed in the health dimension (both of natural and human systems), the economic dimension, and the social dimension [37]. Lower-income households are strongly affected by climate-change-related policies, as a certain policy change, i.e., a tax increase, will comprise a much larger part of their income [36]. Moreover, empirical evidence has shown that lower-income households may have less access to the EU funds that support energy efficiency in comparison with higher-income households [38]. Therefore, socioeconomic inequalities may be amplified.

On the regional side, Rodríguez-Pose and Bartalucci [39] identify two channels, one direct and one indirect, through which the green transition may exaggerate regional inequalities. The green transition will have significant direct impacts in regions dependent on “brown” energy production. Many of these regions are laggards and job and income losses will further deteriorate their position. In addition, carbon-related taxes will mostly harm lower-income households. On the other hand, the transition will indirectly harm lagging regions by increasing factor mobility and reallocating economic and social assets to more advanced economies. This is likely to occur, because green technologies, employment, and innovation are concentrated in areas that are better equipped to absorb capital and labor for sustainable economic activities.

The EU Green Deal is expected to have a major impact on jobs, livelihoods, working conditions, and skills. Even so, a bold reference to “inequality” or “equality” is missing from the EU Green Deal; it only makes a reference to “leaving no-one behind” [40].

Regarding the smart transition, empirical studies emphasize the importance of local conditions and individual characteristics in shaping its effects on spatial and income inequalities. Capello et al. [41] found that the relationship between Research, Technology Development, and Innovation (RTDI) funding under CP and economic growth is complex and context dependent. During recovery periods, this relationship can even turn negative in regions that are more vulnerable to macroeconomic shocks, potentially deepening regional divides. This may be due to the spatial concentration of innovation activities and the uneven ability of regions to capitalize on innovation. A recent study regarding the use of e-government services in Spain revealed that almost half of the population does not use these services; this is mainly because of their digital skills, but is also impacted by their lack of trust in the Internet [42]. This forefronts the need to make public digital skills policies that are inclusive and accessible to all citizens, not just the digitally literate.

Similarly, Lee [43] highlights that the outcomes of innovation policies vary significantly depending on regional contexts. Top–down investments tend to be more effective in areas with an established base of innovative firms. Moreover, innovation activities have been shown to be linked with inequality patterns and present differences among EU Member States; areas in Southern and Eastern Europe show higher inequality while also having the least innovative activities [44]. In addition, Celli et al. [45] show that Cohesion Policy funds allocated to R&D do not necessarily produce stronger economic growth compared to other types of interventions.

Table 2. The findings of key studies on the relationship between smart and green contexts and their dynamics, as well as the impacts of policies supporting the twin transition on inequalities.

Source	Analysis Description	Green/Smart Variables	Inequality Variables	Findings
Green Policy
Tonin [38]	Qualitative research	Energy efficiency measures in the residential sector (tax deduction)	Social and economic wellbeing and mitigation of GHGs	These measures are highly beneficial for societal wellbeing and for reducing energy consumption
Rodríguez-Pose and Bartalucci [39]	Analytical framework (composite measure) (most recent data, ranging from 2018 to 2022)	Regional Green Transition Vulnerability Index	Regional GDP per capita	Negative correlation (poorer regions are more vulnerable to the green transition)
Bettarelli et al. [46]	Econometric analysis—39 developed and developing countries (1990–2020)	Climate change policies (CCP)	Income inequality	CCPs increase income inequality
da Silva et al. [47]	Meta-analysis of 195 different EU policy documents on the European Green Deal	EU Green Deal	Spatial and economic inequalities	Despite ambitious claims about the ethical merits of the transition they aim to bring about, their implementation depends on the extraction of massive amounts of raw materials, especially in Southern EU countries
Smart Policy
Capello et al. [41]	Econometric analysis—EU27 NUTS2 regions and UK (2008–2019)	Cohesion Policy funds	Compound growth rate of gross regional product per capita	Positive relationship: can turn negative in regions more vulnerable to macroeconomic shocks, in times of economic recovery
Caragliu and Del Bo [48]	Econometric analysis—106 EU cities	Urban Smartness	Gini	Negative relationship
Green context
Yamamura [9]	Econometric analysis—86 countries (1965–2004)	Natural disasters	Income inequality	Increase in the short term; in the long term, the impact disappears
Paglialunga et al. [6]	Econometric analysis—150 countries (2003–2017)	Temperature rise and precipitation	Within-country inequality	Increases especially in areas where many people live in rural areas and work in agriculture
Cevik and Jalles [13]	Econometric analysis—158 countries (1995–2019)	Climate change vulnerability	Income inequality	Positive relationship in the case of developing countries
Mejean et al. [16]	Systematic literature review (127 papers)	Climate change impacts	Income inequality	Climate change increases income inequality
Smart Context
Lee [49]	Econometric analysis—EU regions (1995–2001)	Patents	Gini, Theil, Atkinson, P8020	Positive relationship
Antonelli and Gehringer [50]	Econometric analysis—39 industrialized cities, 1995–2011	Number of patent applications made	Gini	Negative relationship (inequality reducing effects start at a certain level of income inequality)
De Palo et al. [44]	Econometric analysis—EU27 and UK	Patents	Gini	Negative relationship
Robles et al. [42]	Qualitative research and correlation analysis—Spain (2018)	Digital Skills	Gender (women)	Negative relationship
Consoli et al. [24]	Regression analysis using 103 EU regions for the period 2003–2013	Regional digital skills endowment	Gini within regions	Increasing e-skills might harm the people at the lower levels of income distribution as they typically possess lower skills

summarizes the findings of key studies on the relationship between smart and green contexts and their dynamics, as well as studies on the impacts of policies supporting the twin transition on inequalities.

The preceding analysis underscores the multifaceted nature of the twin transition and its wide-ranging implications for social welfare and inequality across different spatial levels. For instance, the evidence shows that climate change tends to amplify inequalities, as most of the studies reviewed point in that direction. By contrast, the effects of digitalization and the policy framework appear more complex. It also becomes evident that the majority of studies concentrate either on the dynamics of the transition or on the policies supporting them. Moreover, the smart dimension of the transition seems to dominate research on the impact of the transition on EU cohesion, with increasing attention being devoted almost exclusively to this aspect [51]. Therefore, a more holistic analysis is required to gain a clearer understanding of the twin transition’s implications for both spatial and socioeconomic inequalities.

Thus, this paper aims to fill that gap by examining the interplay between transition dynamics and policy interventions, assessing their joint impacts on inequality at both national and regional levels. The paper quantifies the context and dynamics of smart and green transitions by interchangeably examining two indicators for each side of the transition. Moreover, policies are quantified by examining the share of ESIF funds going to smart and green transition. In addition, the study employs indicators that capture both spatial and income inequalities, providing a comprehensive view of how the twin transitions shape cohesion across the EU. Finally, quantifying the intensity of CP interventions in both smart and green domains, these findings also add to the broader discussion on the effectiveness of the CPs.

The following analysis builds on the elaboration of the following hypotheses:

Smart and green transition dynamics affect spatial and socioeconomic inequalities.

The policy framework supporting smart and green transitions affects spatial and socioeconomic inequalities.

3. Materials and Methods

To address the presented research questions and perform an econometric analysis regarding the effect of green and smart transitions on inequality, the basic variables, their descriptions, and their sources are presented in

Table 3. The variables of econometric analysis.

Variable	Description	Source
Dependent
CVw	Weighted coefficient of variation (NUTS II) of the Regional GDP pc	Eurostat [52]
Gini	Gini index (income)	WID [53]
Explanatory Variables
Smart Transition
Computer_Use	Computer use within last 3 months (% of individuals).	Eurostat [52]
Skill	Share of employment participating in the following highly specialized manufacturing activities: chemicals and chemical products [C20]; basic pharmaceutical products and pharmaceutical preparations [C21]; rubber and plastic products [C22]; computer, electronic, and optical products [C26]; electrical equipment [C27]; machinery and equipment n.e.c. [C28]; motor vehicles, trailers, and semi-trailers [C29]; other transport equipment [C30]	Eurostat [52]
Smart Policy
Smart_sShare	Share of total structural funds going to smart interventions	European Commission [54,55]
Green Transition
CO2pc	CO2 emissions (metric tons per capita)	World Bank [56]
Renewable	Renewable energy consumption (% of total final energy consumption)	World Bank [56]
Green Policy
Green_Share	Share of total structural funds going to green interventions	European Commission [54,55]
Controls
Ln_GDPpc	Natural logarithm of GDP per capita in constant 2015 prices	ARDECO [57]
Ln_ESFpc	Natural logarithm of the total annual payments of European Social Fund per capita	European Commission [54,55]
OPEN	World trade of products and services as a GDP share	World Bank [56]
Ln_Pop_Dens	Natural logarithm of population density (residents per 1 km2)	Eurostat [52]
Inv_Infr_share	The share of structural funds going to investments on infrastructure	European Commission [54,55]
Bus_Tax	Business tax (% of total)	Eurostat [52]

Sources: [52,53,54,55,56]; own elaboration.

. Three variables act as dependent variables ones, quantifying spatial inequality (weighted coefficient of valuation—CVw) and socioeconomic inequality, as measured by the Gini index.

The CVw indicator shows how much variability exists in the allocation of Gross Domestic Product per capita (GDPpc) among the NUTS II regions of a country. Higher values denote a less equal distribution of GDP (see Equation (1)). Since it is weighted, it can better capture the overall dispersion of prosperity among the citizens of a country.

$${CV}_w={\displaystyle \frac{\sqrt{{\sum }_{r=1}^n(X_r-\overline{X}{)}^2{P}_r}}{\overline{X}}}(r=\mathrm{1,2},\dots ,n)$$

(1)

where $x$ is the variable under consideration (GDPpc) and P is the proportion of the population of the r region to the total population of the country [58].

The Gini coefficient is a standard measure of income inequality, ranging from 0 to 1, where higher values indicate greater income concentration [53].

The key explanatory variables cover the dynamics and policies of the smart and green transitions, as presented in Table 3. Both the contextual and policy dimensions of the twin transition can be quantified through a wide range of variables. For instance, climate-related studies provide valuable data such as the number of extreme weather events and detailed meteorological indicators, while innovation dynamics can be captured through indices such as those included in the European Innovation Scoreboard. In this study, we selected indicators based on the criteria of continuous temporal coverage and availability across all countries, ensuring comparability and consistency of the analysis. Smart transition dynamics are quantified using two variables: “Computer_Use”, which measures the share of the population that used a computer in the past three months, and “Skill”, which measures the smart literacy of the workforce in each country by considering the proportion of the workforce employed in high-skilled jobs. Smart policy is quantified by the variable “Smart_Share”, which reflects the share of ESIF funds allocated to smart-related interventions in each country. Green transition dynamics are measured by the total CO₂ emissions per capita (CO_{2_}pc) and the share of “Renewable Energy” in the total final energy consumption. “Green_Share”, similar to smart policies, is proxied by the share of green-related interventions in the total ESIF budget per country.

As the literature review indicates, there is considerable ambiguity regarding the impact of each of the main variables on spatial and socioeconomic inequalities. This is because the effect of the twin transition on inequalities is influenced by many factors, whose presence or absence in the sample countries may lead to significantly different outcomes. Additionally, our prior knowledge is limited by the uncertain results stemming from the interplay between dynamics and policies in these two areas, a relationship further influenced by external factors in each country. For instance, increased computer use may have a different effect on reducing inequalities depending on the level of technological literacy in a given region [59]. The green transition is similarly shaped by the unique characteristics of each country.

The literature review highlighted that climate change tends to mostly harm lagging regions and their populations. However, the transition to a more environmentally friendly energy mix might also negatively impacts regions that derive a substantial portion of their GDP from brown energy activities [60]. Therefore, the final effect of the green transition will be determined by the opposing forces of the shift to green energy and the policies that ensure that this transition is for the good of people and places.

Regarding the controls used in the study, these concern several factors that have been shown in the literature to influence a country’s ability to effectively manage the twin transition while reducing inequalities. The first control variable is the development level of each country, expressed as the total annual GDP per capita in constant 2015 prices. There is substantial research on the relationship between GDPpc and inequality. Following the seminal work of Kuznets, in which the inverted U-shape between economic development and inequalities was documented, many scholars have found that there is a nonlinear relationship between the two phenomena [61]. Nevertheless, despite the wide perception that inequalities diminish with the level of development, there is notable evidence that more advanced countries experience high inequalities, particularly in spatial terms [27,62,63]. On the other hand, the literature is not so enlightening on how the development level of a country affects the impact of green and smart transitions on inequalities due to the role of other structural characteristics. Therefore, richer countries with a high reliance on brown energy may face greater challenges in securing just transitions [56,57], and countries with a high specialization in the technology-related sector may see incomes concentrating to more high-skilled workers and dynamic regions [40,64].

The second control (Ln_ESF_pc) relates to the amount of per capita funding directed to each country by the European Social Fund. The ESF directs finds directly to combating socioeconomic and spatial inequalities; therefore, these funds are expected to have a counteracting effect on inequalities. Moreover, since these funds are provided under detailed Partnership Agreements, a greater portion is expected to support the EU’s broader strategies, especially in lagging regions. Therefore, a strong supporting financial net may both help countries to counteract inequalities and make the twin transition more effective and inclusive for people and places [33,65]. The OPEN variable, measured as the ratio of total trade of goods and services to GDP, quantifies the openness of EU economies, which may influence their ability to adapt to an ever-changing environment and secure more inclusive development among the winners and losers of trade [66,67]. The direction of the effect is ambiguous depending on the ability of places and people to cope with globalization impacts.

Moreover, the concentration dynamics of each country are anticipated to affect their ability to achieve an equitable allocation of outcomes, especially in a spatial context [68]. Agglomeration dynamics are quantified by the population density variable (Ln_Pop_Dens), expressed as the number of residents per sq km in logarithmic terms. The anticipated effect of population density on economic and spatial inequalities is theoretically ambiguous. On the one hand, a greater concentration of the population can lead to the implementation of benefits such as better labor market matching, economies of scale in production, better access to public services such as health and education, lower per capita infrastructure costs, and knowledge spillovers. On the other hand, population density can reinforce existing inequalities in cases where access to the aforementioned benefits is uneven among people and places [69,70]. In addition, the Inv_Infr_share variable quantifies the proportion of structural funds going to investments in infrastructures, which is a critical factor for securing long-term growth for EU countries by improving access to markets, services, and opportunities in lagging regions and marginalized social groups. Therefore, a negative impact on inequalities is expected in this scenario [27]. Additionally, tax policies are expected to significantly influence the magnitude of socioeconomic inequalities [71]. Here, tax policy is proxied by the share of business taxes in the total tax revenues. The sign of the relationship is also ambiguous; on the one hand, higher business tax rates support redistributive policies, which may lower inequalities, but they could also amplify inequalities by discouraging investment, particularly in peripheral or less-developed regions, where location decisions by mobile capital are more sensitive to taxation. The recent empirical evidence has shown that corporate tax cuts lead to higher inequalities [72,73,74]. Therefore, the type of relationship expected is not clear.

The analysis covers 26 EU countries for which data were available for the period 2007–2020 (Croatia is excluded due to missing data for some variables). Spatial inequality is assessed only in countries with more than one NUTS II region; for this, only 21 countries are covered in this type of analysis. The study period spans 14 years (2007–2020) and encompasses two programming periods for structural funding operations. All variables refer to the national level, and even the variable of spatial inequalities (CVw) uniquely quantifies the level of regional inequalities for each country. The basic model to be estimated is presented in

Table 4. The models of the econometric analysis.

Models	Specification
Basic Smart
1	${Gini}_{i,t}=a+{{\beta }_{1}Skill}_{i,t}+{\beta }_2{Smart\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
2	${Gini}_{i,t}=a+{{\beta }_{1}Computer\_Use}_{i,t}+{\beta }_2{Smart\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
3	${CVw}_{i,t}=a+{{\beta }_{1}Skill}_{i,t}+{\beta }_2{Smart\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
4	${CVw}_{i,t}=a+{{\beta }_{1}Computer\_Use}_{i,t}+{\beta }_2{Smart\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
Basic Green
5	${Gini}_{i,t}=a+{{\beta }_{1}CO2pc}_{i,t}+{\beta }_2{Green\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
6	${Gini}_{i,t}=a+{{\beta }_{1}Renewable}_{i,t}+{\beta }_2{Green\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
7	${CVw}_{i,t}=a+{{\beta }_{1}CO2pc}_{i,t}+{\beta }_2{Green\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
8	${CVw}_{i,t}=a+{{\beta }_{1}Renewable}_{i,t}+{\beta }_2{Green\_share}_{i,t-2}+{\beta }_3{Ln\_GDPpc}_{i,t}+{\beta }_4{Ln\_ESFpc}_{i,t-2}+{\beta }_5{OPEN}_{i,t}+{\beta }_6{Ln\_Pop\_Dens}_{i,t}+{\beta }_7{Bus\_Tax}_{i,t}+{\beta }_8{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
Integrated
9	${Gini}_{i,t}=a+{{{{\beta }_{1}Renewable}_{i,t}+\beta }_{2}Skill}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
10	${Gini}_{i,t}=a+{{{{\beta }_{1}Renewable}_{i,t}+\beta }_{2}Computer\_Use}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
11	${Gini}_{i,t}=a+{{{{\beta }_{1}CO2pc}_{i,t}+\beta }_{2}Computer\_Use}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
12	${Gini}_{i,t}=a+{{{{\beta }_{1}CO2pc}_{i,t}+\beta }_{2}Skill}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
13	${CVw}_{i,t}=a+{{{{\beta }_{1}Renewable}_{i,t}+\beta }_{2}Skill}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
14	${CVw}_{i,t}=a+{{{{\beta }_{1}Renewable}_{i,t}+\beta }_{2}Computer\_Use}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
15	${CVw}_{i,t}=a+{{{{\beta }_{1}CO2pc}_{i,t}+\beta }_{2}Computer\_Use}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$
16	${CVw}_{i,t}=a+{{{{\beta }_{1}CO2pc}_{i,t}+\beta }_{2}Skill}_{i,t}+{\beta }_3{Green\_share}_{i,t-2}+{\beta }_4{Smart\_share}_{i,t-2}+{\beta }_5{Ln\_GDPpc}_{i,t}+{\beta }_6{Ln\_ESFpc}_{i,t-2}+{\beta }_7{OPEN}_{i,t}+{\beta }_8{Ln\_Pop\_Dens}_{i,t}+{\beta }_9{Bus\_Tax}_{i,t}+{\beta }_{10}{Inv\_Infr\_share}_{i,t-2}+{\gamma }_i+{\epsilon }_{it}$

.

In Table 4, i = 1, …, N, indicating the number of countries (CY, EE, LT, LU, LV, and MT are not included in the model of spatial inequalities, as they are one-region countries); t = 1, …, T, indicating the number of years; γ = the unit-specific (individual/country/region) fixed effect that captures the all-time-invariant characteristics of the included countries; ε = the idiosyncratic error term.

The estimation of coefficients proceeds in three steps. First, the effect of smart contexts and policies on inequalities is examined using the Skill and Computer_Use variables, interchangeably with the Smart_Share variable, in models where Gini and CVw are the dependent variables (Models 1–4). Next, the same procedure is applied to green transition variables, using the CO₂pc and Renewable variables interchangeably with the Green_Share variable (Models 5–8). Finally, the integrated models incorporate all pairs of smart and green context variables, while keeping the Smart_Share and Green_Share policy variables in all models (Models 9–16). In all models, the same set of control variables is preserved.

Regarding variable specification, several variables are included with a two-year lag, which applies only to policy-related variables. This lag reflects the anticipated delay between policy interventions and their effects on inequalities. After testing several lag structures, a two-year lag was selected; it provided the best statistical significance for the estimated coefficients. Consequently, Smart_Share, Green_Share, LN_ESFpc, and Inv_Infr_share are entered in all models that are lagged by two years.

The models were tested for heteroskedasticity and autocorrelation to determine the most appropriate estimation strategy. Heteroskedasticity in the residuals of the fixed-effects model was assessed using the command xttest3 in Stata 17 software, while first-order autocorrelation was examined using the xtserial test [75,76]. Both tests indicated violations of the assumptions of homoscedasticity and an absence of autocorrelation. Consequently, a Generalized Least Squares (GLS) estimator for panel data (xtgls Stata command) was employed instead of the fixed-effects Ordinary Least Squares (OLS) estimator. GLS is capable of accounting for heteroskedasticity across panels, autocorrelation within panels, and contemporaneous correlation across panels. Under appropriate assumptions, it provides unbiased, consistent, efficient, and asymptotically normal estimates [77]. The GLS approach is particularly valuable when analyzing international data with different levels of economic development and temporal shocks—as is the case in the current twin transition models [78].

Potential endogeneity between some explanatory variables and inequality measures is addressed through two approaches. First, a 2SLS estimator with clustered errors at the country level was used, employing the lagged values (1–2 years) of each explanatory variable as instruments for both CVw and Gini. After estimation, the Durbin–Wu–Hausman test was applied to determine whether the suspected endogenous regressors were correlated with the error term [79,80]. For all variables, the test results were not significant, indicating that endogeneity was not a concern. The same procedure was repeated for all basic and integrated models (Models 1–16), considering all smart and green context and policy variables, as well as LN_ESFpc and Inv_Infr_share, as potentially endogenous. The F-statistic and significance levels of the Durbin–Wu–Hausman tests are reported in the relevant tables. For all basic models (Models 1–8) and integrated models using Gini as the dependent variable, the null hypothesis of non-endogeneity could not be rejected at the 10% level. For three of the four integrated models with CVw as the dependent variable, non-endogeneity could not be rejected only at the 5% level. Given that individual tests for each variable showed no endogeneity issues, and considering the sensitivity of the test increases with the use of instruments, it is concluded that the GLS estimator is appropriate for analyzing the relationships of interest.

Finally, as a robustness check, Generalized Method of Moments (GMM) estimators were applied to the eight basic models. A double-robust estimator was used, with several lag structures tested for potentially endogenous variables. For the CVw models, 2–3 lags were applied to smart and green context variables (Skill, Computer_Use, CO₂pc, and Renewable), while policy variables (Smart_Share, Green_Share, LN_ESFpc, Inv_Infr_share) were treated as predetermined with 1–2 lags. For Gini models, the availability of more cross-sections allowed the use of 2–4 lags. GMM was applied only to basic models, as the number of explanatory variables in the integrated models did not permit the construction of sufficient instruments [80,81]. Under GMM, most estimated coefficients were not statistically significant, likely due to the relatively small sample, which rendered the models less stable. Results, along with tests for first- and second-order autocorrelation and instrument validity, are presented in Appendix A (

Table A1. The results of the regression analysis of smart transition dynamics and policies on socioeconomic and spatial inequalities under the GMM specification.

Coefficient/Tests	Gini		CVw
	Model 1	Model 2	Model 3	Model 4
L1. Gini/CVw	0.6924 ***	0.6370 ***	0.9711 ***	1.0063 ***
Smart_Share	0.0228	0.0237 *	−0.0145	−0.0080
Skill	0.0434		0.5926 **
Computer_Use		−0.0001		−0.0006 ***
lnGDP_pc	−0.0082	−0.0098	−0.0105	0.0048
ln_ESF_cap	0.0004	0.0004	−0.0009	0.0001
OPEN	−0.0171	−0.0167 *	−0.0182	0.0019
ln_Pop_Dens	0.0010	0.0027	−0.0004	0.0001
Business_Tax	−0.0752	−0.0791	0.0109	−0.0291
Inv_Infr_share	0.0256	0.0123	−0.0142	−0.0161
Constant	0.2492 *	0.2945 *	0.1033	0.0040
Diagnostics
Number of observations	338	338	273	273
Number of groups	26	26	21	21
Instruments	25	25	20	20
Wald chi2(9)	62,655.12	42,214.32	24,038.88	244,557.7
Arellano–Bond test for AR(1) in first differences
z	−2.45	−2.34	−2.89	−2.97
Pr > z	0.014	0.019	0.004	0.003
Arellano–Bond test for AR(2) in first differences
z	1.76	1.75	0.87	0.97
Pr > z	0.078	0.08	0.384	0.333
Sargan test of overid. Restrictions
z	20.31	26.05	13.09	11.4
Pr > z	0.161	0.038	0.219	0.327
Hansen test of overid. Restrictions
z	17.73	16.3	7.68	10.29
Pr > z	0.277	0.363	0.66	0.416

Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table A2. The results of the regression analysis of green transition dynamics and policies on socioeconomic and spatial inequalities under the GMM specification.

Coefficient/Tests	Gini		CVw
	Model 1	Model 2	Model 3	Model 4
L1. Gini/CVw	0.7903 ***	0.6903 ***	0.8924 ***	0.9447 ***
Green_Share	0.0299 **	0.0294	0.0188	0.0113
Renewable energy	−0.0001		−0.0005
CO2capita		0.0009		0.0020 **
lnGDP_pc	0.0001	−0.0043	−0.0096	−0.0088 *
ln_ESF_cap	0.0002	0.0002	−0.0004	0.0002
OPEN	−0.0106	−0.0161	−0.0017	−0.0035
ln_Pop_Dens	−0.0019	−0.0009	−0.0036	−0.0015
Business_Tax	−0.0760	−0.0723	0.0147	0.0202
Inv_Infr_share	0.0345	0.0365	−0.0011	0.0017
Constant	0.1324	0.2137	0.1508	0.0926 *
Diagnostics
Number of observations	338	338	273	273
Number of groups	26	26	21	21
Instruments	25	25	20	20
Wald chi2(9)	45,451.6	40,865.69	18,051.37	42,099.05
Arellano–Bond test for AR(1) in first differences
z	−2.61	−2.36	−2.83	−2.93
Pr > z	0.009	0.018	0.005	0.003
Arellano–Bond test for AR(2) in first differences
z	1.84	1.82	0.84	1.01
Pr > z	0.065	0.069	0.404	0.315
Sargan test of overid. Restrictions
z	29.89	26.36	18.9	19.96
Pr > z	0.012	0.034	0.042	0.03
Hansen test of overid. Restrictions
z	17.74	17.62	10.98	10.6
Pr > z	0.277	0.282	0.359	0.39

Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1.

). The remainder of the paper builds on GLS estimator results, with the GMM findings discussed in the conclusions.

4. Results

4.1. Analysis of Smart and Green Transition Variables

Undoubtedly, not all EU countries exhibit the same readiness to achieve the objectives of the twin transition. The joint analysis of countries’ performances in both the green and smart contexts, as well as their policies supporting the transition, is revealing (see Figure 1a–c). The figures present the average values of the variables in the period 2014–2020. As shown in Figure 1a, there is considerable variability across EU countries in the two green transition indicators. Specifically, countries such as Sweden, Latvia, Lithuania, and Croatia maintain extremely low per capita CO₂ emissions while achieving very high shares of renewable energy in total energy consumption, making the green transition relatively easier for them. In contrast, countries such as Luxembourg, Czechia, Germany, and the Netherlands show low penetration of renewable energy while exhibiting relatively high emissions, making the green transition more challenging.

Figure 1b presents countries’ performance regarding the dynamics of the smart transition. Here, the pattern differs: apart from Germany and, to a lesser extent, Czechia, no country displays exceptional or critically poor outcomes on both indicators. Therefore, all countries face challenges in moving toward a smarter society, highlighting the necessity of using multiple metrics to capture the transition’s multidimensional nature.

Finally, on the policy side (see Figure 1c), the most striking observation is that most countries allocate more funds to the smart transition than to the green transition. This appears reasonable for countries such as Finland and Sweden, which already perform well in the green transition and thus may prioritize digitalization. By contrast, countries such as Luxembourg might have been expected to allocate more resources to green interventions, given their relatively weak performance on both indicators. These findings highlight that countries face different challenges in achieving the twin transition, making it more difficult to design a policy mix that ensures a fair and equitable transition.

The main descriptive statistics of the variables are presented in

Table 5. The descriptive statistics of the considered variables.

Variables	Statistics
Dependent	Observations	Mean	Std. dev.	Min	Max
CVw	273	0.323	0.118	0.144	0.629
Gini	338	0.457	0.042	0.341	0.616
Explanatory Variables
Smart Transition: Computer_Use	338	75.101	13.382	35.280	97.525
Smart Transition: Skill	338	0.048	0.027	0.002	0.123
Smart Policy: Smart Share	338	0.150	0.079	0.000	0.473
Green Transition: CO2 pc	338	7072	3159	2970	22,557
Green Transition: Renewable	338	19.378	11.816	0.220	58.400
Green Policy: Green Share	338	0.104	0.083	0.000	0.393
Controls
ln_ESF_pc	338	2.655	1.554	−4.605	5.204
ln_GDP pc	338	10.037	0.6400	8.644	11.506
OPEN	338	0.897	0.409	0.300	1.813
Ln_Pop_Dens	338	4.666	0.919	2.862	7.384
Inv_Infr_share	338	0.182	0. 133	0.000	0.577
Business_Tax	338	0.346	0.072	0.110	0.490

Sources: see Table 3. Own elaboration.

.

We also present scatterplots illustrating the relationships between the variables of green and smart transition dynamics, policies, and the inequality measures CVw and Gini. Since panel data models are employed, the scatterplots are constructed using the demeaned values of the variables to better capture their associations in a multi-period context. Figure 2 displays the scatterplots with the CVw indicator, while Figure 3 presents those based on the Gini indicator.

In addition,

Table 6. The coefficients and statistical significance of the Pearson correlation analysis among the explanatory variables.

	Computer	Skill	Smart_sh	CO2cap	Renewable	Green_sh	LnGDP_pc	LnESF_pc	OPEN	Bus_Tax	Inv_Inf_sh	Ln_Density
Computer_Use	1.00
Skill	0.19 *	1.00
Smart_Share	0.11 *	0.22 *	1.00
CO2pc	0.33 *	0.09	−0.04	1.00
Renewable	0.28 *	0.09	0.04	−0.32 *	1.00
Green_Share	−0.51 *	−0.06	−0.13 *	−0.18 *	−0.21 *	1.00
LnGDP_pc	0.67 *	−0.06	0.08	0.47 *	0.02	−0.64 *	1.00
LnESF_pc	−0.14 *	0.08	0.12 *	−0.05	−0.09	0.07	−0.23 *	1.00
OPEN	0.16 *	0.36 *	0.06	0.11 *	−0.19 *	0.21 *	−0.28 *	0.10	1.00
Bus_Tax	−0.37 *	0.07	0.03	−0.12 *	−0.01	0.34 *	−0.47 *	0.19 *	0.20 *	1.00
Inv_Infrs_share	−0.56 *	0.04	−0.06	−0.24 *	−0.16 *	0.64 *	−0.77 *	0.21 *	0.20 *	0.47 *	1.00
Ln_Density	−0.01	−0.06	0.10	0.14 *	−0.71 *	0.09	0.19 *	−0.02	0.17 *	−0.30 *	−0.09	1.00

Statistical significance: marked with * when estimation is significant at least at the p < 0.05 level.

reports the results of the Pearson correlation analysis among the explanatory variables of the models. The highest correlation is observed between Ln_GDP_pc and Inv_Infr_share. However, the value does not exceed 0.80, which is typically used as a threshold for signaling potential collinearity concerns. To further assess multicollinearity, a pooled regression model was estimated using Ordinary Least Squares (OLS) with CVw as the dependent variable, including all the explanatory variables. The Variance Inflation Factor (VIF) values were then calculated. The highest VIF was found for Ln_GDP_pc (5.28), while no other variable exceeded a VIF of 5. The mean VIF was 2.79. Given that the critical threshold for VIF is conventionally set at 10, these results indicate that multicollinearity is not a problem in this dataset [82].

4.2. Results of the Econometric Analysis

Following the graphical representation of the key variables in relation to inequality, and the correlation analysis among the explanatory variables, the results of the models on spatial inequalities and the smart transition are presented in

Table 7. The results of the regression analysis of smart transition dynamics and policies on socioeconomic and spatial inequalities.

Coefficients/Tests	Gini		CVw
	Model 1	Model 2	Model 3	Model 4
Skill	−0.388 ***		0.217 *
Computer_Use		−0.0004 **		0.0003
Smart_Share	−0.002	−0.007	−0.005	−0.021 ***
Ln_GDPpc	−0.034 ***	−0.017 ***	−0.106 ***	−0.110 ***
Ln_ESFpc	0.0006	−0.0005	−0.0002	−0.0010 ***
OPEN	−0.021 **	−0.015 **	0.021 ***	0.016 ***
Ln_Pop_Dens	−0.007	−0.007	0.005	0.012 *
Bus_Tax	−0.181 **	−0.198 **	0.073 ***	0.121 ***
Inv_Infr_share	−0.021 **	−0.013	−0.011	−0.010
_cons	0.935 ***	0.770 ***	1.310 ***	1.299 **
Diagnostics
Wald	133.270 ***	76.160 ***	815.790 ***	1130.590 ***
Endogeneity Test
F	0.775	0.806	1.970	1.286
p	0.552	0.533	0.138	0.309

Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1.

. Four models with different specifications are estimated, with all models including the core independent variables while successive specifications progressively add more control variables and interaction terms. Across all models, the Wald test indicates a good overall model fit, while the endogeneity tests consistently fail to reject the null of exogeneity at high levels of statistical significance.

The estimated coefficients reveal meaningful but nuanced differences across models. Both context variables—Skill and Computer_Use—exhibit a negative relationship with the Gini coefficient, suggesting that smart transition dynamics may reduce socioeconomic inequalities within countries. However, the intensity of smart transition policies (Smart_Share) does not show a systematic effect on socioeconomic inequalities. By contrast, in the case of spatial inequalities, only Skill shows a significant association, with a positive coefficient indicating that a larger share of skilled labor may exacerbate regional disparities, in contrast to its mitigating effect on socioeconomic inequalities. This result, however, is only significant at the 10% level and should therefore be interpreted with caution.

At the policy level, Smart_Share exhibits a negative and statistically significant effect on spatial inequalities in Model 4 (p < 0.01), suggesting that smart policies may play a role in reducing territorial disparities. Regarding the control variables, national income (Ln_GDP_pc) consistently shows a negative and highly significant association (p < 0.01) with both socioeconomic and spatial inequalities, underscoring the inequality-reducing role of economic development.

Two particularly noteworthy findings concern OPEN (trade openness) and Bus_Tax (business tax). Both variables display negative and statistically significant coefficients in the models for socioeconomic inequality, implying that trade integration and higher corporate taxation may enhance social cohesion. However, for spatial inequality, both variables exhibit positive and significant coefficients, indicating that these factors may exacerbate territorial disparities. This dual effect highlights the complex policy challenge of addressing both forms of inequality simultaneously.

Finally, EU structural funds (Ln_ESFpc) are found to reduce spatial inequalities in one model specification, while investment in infrastructure (Inv_Infr_share) appears to reduce socioeconomic inequality, though its effect is weaker and significant only in one specification. Population density does not show a consistent effect, with only one marginally significant coefficient (p < 0.10).

In addition,

Table 8. The results of the regression analysis of green transition dynamics and policies on socioeconomic and spatial inequalities.

Coefficients/Tests	Gini		CVw
	Model 5	Model 6	Model 7	Model 8
CO2pc	0.002 *		0.003 **
Renewable		0.00002 *		−0.003 ***
Green_Share	−0.032 **	−0.022	−0.043 ***	−0.043 ***
Ln_GDPpc	−0.037 ***	−0.028 ***	−0.112 ***	−0.097 ***
Ln_ESFpc	−0.0004	−0.0001	−0.0004	−0.0002
OPEN	−0.030 ***	−0.032 ***	0.024 ***	0.016 *
Ln_Pop_Dens	−0.006 *	−0.007	0.010	−0.024 ***
Bus_Tax	−0.194 ***	−0.196 ***	0.086 ***	0.056 ***
Inv_Infr_share	−0.015	−0.020 **	−0.016 *	−0.020 ***
Constant	0.949 ***	0.877 ***	1.334 ***	1.441 ***
Diagnostics
Wald	102.250 ***	151.110 ***	1015.520 ***	879.220 ***
Endogeneity Test
F	1.310	0.782	0.868	1.556
p	0.294	0.548	0.500	0.225

Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the results of the basic models on the green transition. Compared with the smart transition analysis, the findings are more contrasting. Both green context variables appear to amplify socioeconomic inequalities, although their coefficients reach statistical significance only at the 10% level. For spatial inequalities, CO₂ emissions per capita also display a positive association, indicating an inequality-enhancing effect. By contrast, the renewable energy consumption share shows a negative and statistically significant coefficient, suggesting that greater penetration of renewable energy helps reduce regional disparities. This result highlights the urgency of reducing emissions, while also underscoring that expanding renewable energy use can provide lagging regions with opportunities to diversify economic activities. However, policies must ensure that access to renewable energy is widely available, so that these benefits do not come at the cost of weakening overall social cohesion.

Furthermore, the Green_Share variable—representing the policy dimension—shows a consistently negative and statistically significant relationship with inequalities in three of the four model specifications. This suggests that allocating more funds to green transition interventions may be effective in reducing both income and regional disparities. The control variables behave similarly to those in the smart transition models, with one notable exception: population density is found to have a negative and strongly significant effect in one of the models of spatial inequality. This finding suggests that density should not be considered a priori as a risk factor for territorial cohesion, since its effect likely depends on complementary structural and policy factors.

Finally, the results of the integrated models are presented in

Table 9. The results of the regression analysis of green and smart transition dynamics and policies on socioeconomic and spatial inequalities.

Coefficients/Tests	Gini				CVw
	Model 9	Model 10	Model 11	Model 12	Model13	Model 14	Model 15	Model 16
Renewable	0.00004	0.0002			−0.004 ***	−0.003 ***
CO2_pc			0.003 ***	0.002 **			0.001	0.002
Skill	−0.369 ***		−0.444 ***		0.505 ***		0.414 ***
Computer_Use		−0.001 ***		−0.0003 *		0.001 ***		0.0005 ***
Green_Share	−0.015	−0.010	−0.012	−0.029 ***	−0.057 ***	−0.026 ***	−0.028 ***	−0.031 ***
Smart_Share	−0.003	−0.003	0.006	−0.005	−0.010 **	−0.008 ***	−0.022 ***	−0.030 ***
Ln_GDPpc	−0.032 ***	−0.021 **	−0.037 ***	−0.028 ***	−0.106 ***	−0.113 ***	−0.106 ***	−0.116 ***
Ln_ESFpc	0.0000	−0.0005	−0.0002	−0.001 *	−0.0004	−0.001 ***	−0.001 ***	−0.001 ***
OPEN	−0.020 **	−0.022 ***	−0.023 ***	−0.023 ***	0.010	−0.003	0.008	0.017 ***
Ln_Pop_Dens	−0.010	−0.003	−0.008 **	−0.006	−0.034 ***	−0.017 ***	0.014 **	0.011
Bus_Tax	−0.194 ***	−0.182 ***	−0.154 ***	−0.185 ***	0.073 ***	0.065 ***	0.138 ***	0.163 ***
Inv_Infr_share	−0.003	−0.018 **	−0.010	−0.017 ***	−0.032 ***	−0.018 ***	0.001	−0.009
Constant	0.932 ***	0.808 ***	0.944 ***	0.868 ***	1.568 ***	1.520 ***	1.252 ***	1.332 ***
Diagnostics
Wald	130.44 ***	107.18 ***	230.90 ***	115.43 ***	1712.35 ***	1121.58 ***	684.73 ***	689.46 ***
Endogeneity Test
F	0.882	0.895	1.360	1.555	2.489	1.760	2.580	2.150
p	0.522	0.514	0.269	0.202	0.058	0.159	0.051	0.092

Statistical significance: *** p < 0.01, ** p < 0.05, * p < 0.1.

. For the green dimension of the transition, the findings validate the detrimental effect of increased CO₂ emissions on social cohesion, as the coefficient of the CO₂_pc variable was positive and statistically significant in both Gini models. However, emissions appear to play no role in shaping spatial inequalities. By contrast, the greater penetration of renewable sources in the energy mix seems to reduce spatial economic disparities, with the coefficient of Renewable_share found negative and statistically significant in both CVw models. Nevertheless, the integrated models do not confirm the positive relationship between renewable energy share and socioeconomic inequalities observed in the earlier results, since the coefficients, although positive, lack statistical significance. On the policy side, green interventions appear to influence primarily spatial inequalities, as indicated by the statistically significant negative effect of Green_Share on the CVw indicator. In the context of the energy transition, such policies seem particularly important, as they can counterbalance the disruptive effects of decarbonization in previously industrialized regions. Regarding socioeconomic inequalities, the variable retains a beneficial (negative) effect, although statistical significance was observed only in one specification (Model 12). Taken together, these findings underline the value of strengthening green policy interventions as a means of promoting both social and territorial cohesion across EU countries

Turning to the smart transition dimension, the results on the contextual variables are more complex, as their effects vary by the type of inequality considered. Specifically, both Computer_Use and Skill show a negative and statistically significant effect on socioeconomic inequalities (Models 9–12), while the effect is reversed when spatial inequalities are considered (Models 13–16). This duality indicates that smart transition dynamics can simultaneously reduce income inequality while exacerbating regional disparities. The smart policy variable, Smart_Share, consistently exhibits negative coefficients across all models, although statistical significance is attained only under the CVw specifications. As with green policies, these results suggest that intensifying smart interventions may be particularly promising for reducing spatial inequalities.

Regarding the control variables, the level of economic development (Ln_GDPpc) is negatively related to both types of inequality, with coefficients consistently significant across all specifications. ESF funding also emerges as an important factor for counteracting regional disparities, as the respective coefficients are negative and statistically significant in three of the four spatial inequality models (Models 14–16). For income inequalities, ESF funds also show a negative effect, though statistical significance (at the 10% level) is found only in Model 12. The dual effects of trade openness (OPEN) are again evident, although the inequality-amplifying effect is statistically significant only for one spatial inequality model this time. By contrast, the Bus_Tax variable provides stronger evidence of opposing effects, showing a negative and statistically significant impact on income inequality but a positive and significant effect on spatial disparities across all specifications. Investment in infrastructure also emerges as a potentially equalizing factor, with negative and statistically significant coefficients in four of the eight models, pointing to its capacity to reduce both income and regional inequalities. Finally, the role of population density (Ln_Pop_Dens) appears to be sensitive to model specification. While the variable carries a negative coefficient in all income inequality models (significant in one case, suggesting that population concentration may help reduce income disparities), its effect on spatial inequalities depends on the green contextual variable considered. In the models where CO₂_pc is included, the coefficient changes sign; in one case (Model 13), it is positive and statistically significant, raising further questions about the real impact of concentration on territorial cohesion.

5. Discussion

The paper sheds light on the effects that a smarter and greener policy mix may have on economic and spatial inequality. The main contribution of the present paper to the relevant literature is that it tested the effect of the dynamics of the smart and green transition and the effects of the policies supporting them on spatial and economic inequalities under a holistic approach. That is, both facets of the twin transition were quantified and incorporated into the analysis, with the aim of revealing their impact on both spatial and income inequalities in EU countries.

The integrated analysis of smart and green transitions highlighted several important dynamics in shaping socioeconomic and spatial inequalities across EU countries. For the green dimension, CO₂ emissions were found to significantly exacerbate income inequality, though they had no detectable effect on spatial disparities. In contrast, a higher share of renewable energy in the energy mix contributed to reducing regional disparities, while its effect on income inequality was not robustly supported. At the policy level, green transition interventions consistently reduced spatial inequalities and showed some beneficial effects on income inequality, though statistical significance for the latter was weaker. These results stress the importance of strengthening green policies, particularly as tools to offset the disruptive impacts of decarbonization on previously industrialized regions.

For the smart transition, the results were more complex. Both the Computer_Use and Skill variables reduced income inequality, with effects that were statistically significant across multiple specifications. However, the same variables were found to exacerbate spatial inequalities, suggesting that, while smart dynamics may foster social cohesion at the national level, they risk amplifying disparities between regions. Smart policies (Smart_Share) generally pointed in the direction of reducing inequalities, especially spatial disparities, although their effects were statistically significant only under specific model specifications. This duality underlines the need for nuanced policy design, as the smart transition can simultaneously deliver inclusive and divisive outcomes depending on the dimension of inequality considered.

Control variables played an important role in validating the robustness of these findings. Economic development (Ln_GDPpc) consistently reduced both income and spatial inequalities, with coefficients significant across all specifications, reaffirming the central role of growth in fostering cohesion. EU funding through the ESF also proved especially effective in reducing regional disparities, with strong statistical support in most spatial inequality models, while its role in mitigating income inequality was weaker. Infrastructure investment also stood out as a potentially equalizing factor, significantly reducing both types of inequalities in several models. By contrast, the effects of trade openness and business taxation were less clear, with both showing beneficial impacts on income inequality but reinforcing spatial disparities.

Overall, the strength of the results varies across models’ specifications. Findings on the role of economic development and ESF funds are particularly robust and consistently supported across model specifications. Results for contextual variables such as CO₂_pc, Skill, and Computer_Use are also statistically meaningful, though their effects differ by inequality type. Policy variables, while generally pointing in the direction of reducing disparities, exhibit weaker significance in some cases, suggesting that their influence may be more context-dependent. In conclusion, the findings demonstrate that the green and smart transitions can play a dual role: they hold potential for reducing inequalities but can also exacerbate them if policies are not carefully tailored to address both social and spatial dimensions simultaneously.

Finally, a few words should be said about the results of the GMM estimator models (see Appendix A). The diagnostic tests indicate that the selected instruments are mostly valid in the Gini models, as the Arellano–Bond test for second-order autocorrelation was not statistically significant, confirming the absence of problematic serial correlation in the differenced residuals. By contrast, in the CVw models, the null of no second-order autocorrelation could only be rejected at the 5% level, raising concerns about the validity of the instruments in this specification. It is important to note that statistical significance in the AR(1) test is expected due to the first-differencing effect, while the lack of significance in the AR(2) test is a necessary condition for instrument validity in dynamic GMM panel.

Moreover, the limited number of observations was constrained the number of lags that could be used, which may partly explain the lack of statistically significant estimates for most variables included in the analysis. For comparative purposes, the GMM results partly confirm the positive effects of the Skill and CO₂ emission variables, as well as the negative effect of the Ln_GDPpc variable on spatial inequalities. However, the models also returned a positive effect of the Green_Share and Smart_Share variables on the Gini coefficient, although this was only significant at the 10% level. Finally, an unexpected and strongly statistically significant negative effect of the Computer_Use variable on spatial inequalities was found, which did not appear in the GLS specifications.

6. Conclusions

The preceding analysis highlights the value of a holistic approach in studying the twin transition, integrating both green and smart dimensions alongside multiple measures of inequality. By considering socioeconomic and spatial disparities simultaneously, the study provides a more nuanced understanding of the way in which transition dynamics and related policies interact across countries. This comprehensive perspective is crucial, as it reveals the dual and sometimes opposing effects of key drivers—such as skill development, technology adoption, and smart policies—on different dimensions of cohesion. That is, the different dimensions of the two transitions may have varying effects on socioeconomic and spatial inequalities. This approach highlights the importance of prioritizing more holistic approaches on the issue.

Moreover, the findings have strong policy implications. The results highlight that interventions targeting smart or green transitions cannot be designed in isolation: actions that may promote socioeconomic equality could exacerbate spatial disparities, and vice versa. The results therefore emphasize the importance of coordinated, multidimensional policies that account for both social and regional impacts, ensuring that the benefits of the transition are equitably distributed across all regions and segments of society.

EU policy places strong emphasis on enhancing the competitiveness of its countries in a rapidly evolving global context. At the same time, emerging mega-trends, such as the digital revolution and climate change, will continue to add pressure on those places and people that are being left behind. Therefore, boosting competitiveness requires an effective twin transition in the EU. This path will have implications for social cohesion, since the empirical evidence suggests that both market dynamics and policy design related to the twin transition affect inequality. Unfortunately, the road to a smarter and greener Europe is not without challenges, the most important being to ensure that this goal will not harm the EU’s social and spatial cohesion. This challenge is particularly important, as countries enter the transition process with different starting points and background conditions, as highlighted in the introductory analysis of the Results Section. A promising finding from the present analysis is that both green and smart policy interventions appear to have counteracting effects on inequalities.

Nevertheless, in order for smart and green policies to retain their effectiveness in the field of inequalities, they should be tailored to the specific economic and industrial contexts of each region. In implementing smart interventions, as stressed at the EU Commission meeting in July 2020, it is important to note that a reduction in skill inequality affects both human capital and social and structural equality [83]. Thus, EU policymakers need to focus both on investments in new technologies and on supporting weak regions in terms of digital equality [3]. In the green dimension, investments in green energy should be accompanied by measures that support economic diversification in regions that are dependent on carbon-intensive industries. Recent empirical studies show that the progress of EU countries in terms of adopting Green Deal provisions [84] varies heavily; policies that work well in one country may be ineffective in others. This suggests that policies need to be tailor-made on the basis of national realities and capabilities; it is important that we avoid the broad implementation of ineffective, one-size-fits-all innovation policies [43]. Realizing the Green Deal’s vision requires a unified implementation strategy, which will include green innovations, substantial investments, and adaptive regulatory frameworks to enhance the energy transition in left-behind regions [85]. The Just Transition Fund is a promising initiative in this context, but it should be implemented along with other, complementary measures from other initiatives [86]. This fund is directed towards the highly dependent coal, lignite, oil shale, and peat industries, assisting them in reskilling their affected workers and unemployed people [87].

Moreover, greater emphasis should be given to monitoring the impact of the twin transition on all types of inequalities. Cross-disciplinary collaboration should be established in order to better monitor and adjust policy mixes based on inequality impacts. Furthermore, strengthening institutional support is crucial, as effective institutions manage to handle green and smart transitions in a way that promotes social cohesion. Ensuring that public investment is directed towards those regions that are in the most need, and ensuring that investments are managed effectively, will be critical in mitigating the inequality-enhancing effects of green and smart transitions.

Finally, it should be noted that the present paper assessed the effect of twin transition on inequalities, considering only the part of relevant policies encompassed in the CP framework. As the transition progresses, other funding mechanisms are at play that might affect social and regional cohesion. Therefore, the present analysis could be enriched with model specifications that also consider these drivers. As stated earlier, the aim of this paper was to develop an integrated framework to test the effects of both green and smart transitions on inequalities. However, the inclusion of a wide range of variables comes at the cost of reduced flexibility in addressing potential endogeneity issues with suitable instruments. In this respect, a larger sample—using regions as the unit of analysis—could enable a better specification of models with GMM estimators, thereby allowing for a more robust treatment of endogeneity. Endogeneity could also be treated with a more focused approach on specific dimensions of the twin transition. Furthermore, given the already dense specification of the models employed in this study, nonlinear relationships were not explored. Expanding the sample size could also facilitate the investigation of such nonlinear effects. The inclusion of additional data covering the current programming period would be very beneficial in our move in this direction. Finally, the variables used here to quantify transition dynamics could be complemented by other measures of the two phenomena.