Discourse vs. Decarbonisation: Tracking the Alignment Between EU Climate Rhetoric and National Energy Patterns

Abstract

This study examines the alignment between the European Union’s climate policy rhetoric and the actual fossil fuel consumption behaviours of its Member States. By combining long-term and short-term time-series data with machine learning classification techniques, the analysis captures dynamic national energy trends and decarbonisation signals. Key innovations include the use of slope acceleration metrics and the identification of label reversals to detect volatility, acceleration, or stagnation in transition trajectories. The results show that, while some countries such as France and Denmark demonstrate consistent structural progress, others show deceleration or reversal, particularly in the use of gas and liquid fuels. This indicates that the relationship between EU-level policy ambition and national implementation is asymmetric and conditionally aligned. This study concludes that ongoing empirical monitoring and targeted diagnostics are essential to prevent conflating symbolic commitments with material change, and provides practical insights for improving climate policy accountability and adaptability across the EU.

1. Introduction

The European Union (EU) has established itself as a global leader in climate policy, supported by bold frameworks such as the European Green Deal and the Fit for 55 package. These initiatives outline a clear normative vision for swift decarbonisation, seeking to cut greenhouse gas emissions, speed up the adoption of renewable energy, and phase out fossil fuels. Nonetheless, the extent to which these political commitments lead to tangible behavioural and structural changes at the national level remains a vital and insufficiently explored issue.

This study addresses the growing need to empirically assess whether EU Member States are aligning their actual energy transition trajectories with the rhetoric of climate ambition. While much of the policy discourse highlights progress, particularly in setting targets and mobilising green investments, the empirical reality is often more fragmented. National-level energy behaviours, especially patterns of fossil fuel consumption, may not always reflect the transformative narratives promoted at the EU level.

Using machine learning techniques to analyse recent fuel consumption data, this paper examines the extent and consistency of national decarbonisation signals. Beyond identifying general downward trends, the study includes slope acceleration metrics and label reversals to uncover nuanced patterns, such as rebound effects and temporal inconsistencies, which are often missed in policy evaluations. The findings indicate that, while some Member States are indeed reducing their reliance on fossil fuels, others show stagnation or reversal, suggesting a partial and asymmetrical translation of policy into behaviour.

In highlighting the gap between discourse and data, this research contributes to a more grounded understanding of the EU’s energy transition. It underscores the need for continuous, multidimensional monitoring and cautions against conflating aspirational narratives with material outcomes. Such analysis is vital for recalibrating policy instruments, improving accountability, and ensuring that green commitments are not merely symbolic, but structurally embedded.

2. Literature Review

While extensive research has been conducted on the evolution of EU climate governance and the design of its flagship policies, less attention has been paid to the empirical evaluation of how these commitments translate into tangible energy transition behaviours. Existing literature tends to focus either on policy analysis or energy statistics in isolation, creating a methodological and analytical gap. This study aims to bridge this divide by applying machine learning methods to national-level fossil fuel consumption data, thereby contributing to a more integrated understanding of the alignment between rhetoric and reality within the EU energy transition.

2.1. EU Climate Governance and Fossil Fuel Policy Instruments

The European Union (EU) has established a comprehensive framework to regulate climate policies, focusing on reducing fossil fuel consumption to combat climate change and promote sustainable energy solutions. This governance system has developed considerably, characterised by a mix of hard and soft policy tools that involve member states within a multi-level governance structure. This literature review summarises key contributions to understanding the link between EU climate governance and fossil fuel policy instruments [1,2].

Central to the EU’s climate governance are legally binding frameworks and policy mechanisms, such as the Emissions Trading System (ETS) and the European Green Deal (EGD). The ETS, established in 2005 and expanded through subsequent revisions, sets a cap on emissions and allows for trading among member states to incentivise reductions in fossil fuel consumption [3].

According to Knodt, the ETS has become a vital ‘hard governance’ tool, affecting around 45% of overall EU emissions, thus imposing substantial limits on fossil fuel use while encouraging the adoption of renewable energy sources [4,5].

The EGD represents a comprehensive approach to achieving climate neutrality by 2050, setting ambitious targets that shape the energy landscape across EU member states [3].

Furthermore, the EU’s governance framework includes a blend of soft law approaches, increasingly supported by strict reporting and accountability systems. As Schoenefeld and Jordan point out, the EU’s climate policies incorporate aspects of ‘harder’ governance, thanks to developing monitoring and enforcement mechanisms, which improve their effectiveness and the accountability of member states in achieving climate objectives [6].

The inclusion of National Energy and Climate Plans (NECPs), required from each member state under governance regulations, demonstrates how the EU aims for consistency across diverse political landscapes [4,5].

The integration of climate and energy policies through the Energy Union strategy highlights the need for coordinated action among member states. Szulecki and Claes advocate for a better understanding of the overlaps between energy governance and climate objectives, emphasising the role of national policies in shaping EU legislative outcomes [7].

Member states, particularly those with strong economies, such as Germany, have been influential in shaping EU climate ambitions. However, their policies are often reinforced by EU frameworks that encourage cross-border cooperation [8].

Achieving the EU’s climate objectives requires cooperation and alignment among national policies within the wider EU framework. Challenges persist as the EU balances political willingness, energy security, and public acceptance of transformative energy policies. Geopolitical developments, such as the war in Ukraine and the resulting energy crisis, have prompted urgent re-evaluations of Europe’s reliance on fossil fuels.

These developments highlight vulnerabilities in energy supply chains and reinforce the need for the EU to expedite its transition to sustainable energy systems [9,10].

The EU’s capacity to pivot effectively in response to external pressures will be crucial in fulfilling its climate commitments, necessitating resilient governance frameworks that can adapt to dynamic political and economic environments [11,12,13].

2.2. From Policy Rhetoric to Policy Implementation

A key challenge in incorporating long-term views is the natural tension between energy security and climate goals. The EU’s governance system, shaped by the varied interests and sovereignty of its member states, often focuses on immediate issues, such as maintaining energy supply stability, rather than on broad decarbonisation goals. Aalto and Temel emphasise the resistance of member states to ceding control over energy sources and strategies, which leads to fragmented energy policies and hampers coordinated long-term planning.

This situation is exemplified by the ongoing geopolitical considerations that affect EU-Russian energy relations, highlighting how external crises can shift policy focus away from long-term sustainability goals [14].

Furthermore, the governance of energy efficiency remains a crucial part of the EU’s energy policies. The Energy Efficiency Directive emphasises the importance of holding member states accountable for reaching specific energy reduction targets [15].

However, integrating energy efficiency measures into long-term energy strategies often encounters barriers related to under-investment and political opposition, particularly when short-term economic interests conflict with environmental sustainability [16].

Galán-Martín et al. highlight that delays in deploying innovative technologies, such as Carbon Dioxide Removal (CDR), could jeopardise the EU’s long-term climate objectives, emphasising the urgency of immediate action alongside strategic foresight [17].

The integration of new technologies, such as renewable energy sources and energy storage systems, must be strategically planned to align with the EU’s decarbonisation targets. Findings from Oberthür reveal that while current policies may have a short-sighted focus, there is an ongoing discourse aimed at enhancing the long-term viability of these initiatives [18,19].

The push for a comprehensive legal and institutional framework designed to facilitate renewable energy integration demonstrates an effort to go beyond immediate challenges in favour of sustainable long-term development [20].

Moreover, the short-termism common in EU energy and climate policies has been criticised for weakening the effectiveness of long-term strategies. Gheuens and Oberthür contend that a narrow focus on immediate objectives results in policies that may not fully address the science-based requirements for long-term decarbonisation.

This discrepancy highlights the need for policies that not only meet immediate targets, but also adapt to evolving scientific insights on climate change. The complexity of establishing a cohesive energy strategy within the EU is increased by the requirement for coordination among various stakeholders, including national governments, private sector actors, and civil society [21]. The differing levels of commitment and capacity among member states to implement long-term energy policies can hinder progress. For example, member states might prioritise financial stability and energy security over the bold measures necessary for environmental sustainability [22,23,24].

In conclusion, incorporating long-term perspectives into EU energy governance presents significant challenges, ranging from geopolitical factors to the complexities of member state sovereignty and investment priorities. To effectively overcome these obstacles, a concerted effort is necessary to align short-term actions with long-term climate ambitions. Continuous assessment of current policies and frameworks, such as the Energy Efficiency Directive and the European Green Deal, will be vital to ensuring that the EU not only satisfies its immediate energy needs, but also lays the groundwork for a sustainable future.

2.3. Empirical Approaches to Energy Transition Monitoring

In the context of transitioning to renewable energy, the empirical monitoring of energy transition processes is crucial for effective governance and policy implementation. This literature review examines empirical approaches to monitoring energy transitions, with a focus on evaluating frameworks that can accurately capture the complexities and multidimensional aspects of energy consumption and production.

A major challenge, including monitoring energy transitions, is the need for comprehensive evaluation methods that consider multiple influential factors. Gałecka and Pyra advocate for a holistic framework that looks beyond single indicators, emphasising the importance of including a wide range of variables affecting energy consumption patterns [25]. Their analysis responds to the requirements identified by other scholars who argue that monitoring progress in the energy transition requires a multifaceted approach that accounts for various multidimensional variables.

As nations navigate the transitional phase, there is a clear recognition of the ongoing importance of fossil fuels during this period of change. Phan emphasises that, despite the growth of renewables, oil and gas are expected to remain crucial in the short term [26]. This reality requires monitoring systems that not only track the uptake of renewables, but also recognise the continued reliance on traditional energy sources. Innovative methods that allow for the flexible management of energy resources can offer valuable insights into how the transition might influence energy markets and consumption trends over time.

Technological innovation also acts as a crucial component in the transition process. Lukashevych et al. emphasise the need for new technical solutions to manage renewable energy variability [27]. Monitoring frameworks should include metrics that measure technological progress, as these significantly impact the stability and reliability of renewable energy supply. The integration of artificial intelligence (AI) into monitoring offers promising opportunities for improved data analysis and predictive modelling, enabling policymakers to make informed decisions that account for current trends and uncertainties in energy use [27,28]. Equity and justice considerations also remain essential in energy transitions, as highlighted by Jacome et al. [29].

The evaluation of energy justice metrics can illuminate disparities in the distribution of benefits and burdens among communities. Establishing quantifiable goals related to energy equity can enhance accountability and transparency in energy transition policy, ensuring that marginalised groups are not left behind during this pivotal shift. The analytical frameworks proposed by such studies can help policymakers track justice-related outcomes and inform equitable energy initiatives.

Furthermore, comprehensive and targeted instrument mixes are essential for encouraging effective energy transitions. Rosenow et al. emphasise the need for diverse public policy interventions that guide transitions towards climate goals [30]. Their findings indicate that a customised combination of instruments can improve energy efficiency and support the adoption of renewable technologies, facilitating faster transitions. Ongoing monitoring of these policy instruments can offer valuable feedback to help adjust strategies in real-time.

Lastly, accountability mechanisms are integral to the success of energy transition initiatives. Sareen discusses cross-sectoral metrics as tools for accountability in the context of integrating energy systems [31]. Implementing these metrics can foster transparent monitoring and delineate the interactions between energy systems, thereby providing a more integrated view of the transition process.

In contrast to prior literature that has primarily focused on either policy analysis [3,6] or energy statistics [24,30] in isolation, this study contributes a novel framework that explicitly integrates both. We apply machine learning classification to country-level fossil fuel trends and introduce slope acceleration and label reversal diagnostics to identify dynamic, non-linear patterns of energy transition.

This multidimensional approach enables an empirical assessment of whether EU Member States are substantially aligning with their climate policy commitments. As such, the study offers a methodological advancement over prior work, which often assesses compliance using static or cumulative indicators without capturing momentum or reversibility.

3. Materials and Methods

3.1. Research Design and Scope

This study employs a mixed-methods, data-driven approach to assess the degree to which the fossil fuel consumption behaviours of EU Member States align with the bloc’s climate policy rhetoric. The primary objective is to determine whether observable decarbonization trends align with political commitments outlined in instruments such as the European Green Deal and the Fit for 55 packages. The analysis is based on national-level time-series data on fossil fuel consumption, disaggregated by fuel type (gas, liquid, solid), across a sample of EU countries. Two temporal windows are analysed: a long-term historical window (covering the whole available series for each country) and a short-term window (capturing the most recent five-year period). This dual-window design enables a comparative assessment of structural versus recent behavioural changes.

The historical window includes all available data points per country (typically from 2000 onward), capturing long-term structural patterns in fossil fuel consumption. The recent window spans the most recent five-year period (2018–2022), which was chosen to isolate behavioural changes associated with major policy, economic, and geopolitical events, including the European Green Deal, REPowerEU, COVID-19, and the 2021–2022 energy price crisis. This dual-window structure allows for the model to detect both enduring decarbonisation efforts and short-term trend reversals or accelerations. The five years were selected as a compromise between capturing near-term effects and retaining sufficient data for trend calculation. This approach aligns with the current transition monitoring literature, which recommends time-sensitive diagnostics to distinguish between structural and reactive dynamics.

3.2. Data Sources and Preprocessing

The primary dataset comprises annual energy consumption figures (in kilotonnes of oil equivalent, ktoe) obtained from authoritative sources, including Eurostat and national energy agencies (see the Data Availability Section). Fossil fuel consumption is categorised into three main types: Gas (including natural gas and derived gases), Liquid fuels (primarily petroleum and oil products), and Solid fuels (such as coal and lignite). All-time series were normalised and detrended where necessary to reduce baseline bias between countries with different absolute consumption levels.

3.3. Feature Engineering and Trend Diagnostics

For each country and fuel type, a set of quantitative features was derived to capture the shape and dynamics of energy use trajectories: (1) Slope (trend gradient): Linear regression coefficient over each time window, (2) Delta (net change): Total change in ktoe between window start and end, (3) Variance and volatility indicators: Standard deviation of annual changes and (4) Recent acceleration: Difference in slope between recent and historical windows. These features served both as inputs for classification models and for interpretive diagnostics, such as slope differential analysis. These features were selected to reflect different dimensions of energy transition dynamics. Slope captures the overall direction and rate of change (positive, negative, or flat), reflecting the trend trajectory. Delta measures the total net change in ktoe, which helps identify the magnitude of transition regardless of whether the trend is linear. Variance and volatility indicators provide insight into the stability or fluctuations in year-on-year consumption, often linked to economic or geopolitical events. Recent acceleration (ΔSlope) helps detect policy effects or behavioural changes that have emerged only in the short term, by comparing short-term and long-term trends. This combination enables us to assess both the level and momentum of transition efforts, in line with calls for dynamic and multidimensional monitoring frameworks [24,30,31].

3.4. Classification Modelling with Machine Learning

A supervised machine learning approach was employed using a Random Forest classifier, selected for its robustness to nonlinearities and its ability to handle small datasets with high dimensionality. Countries were labelled as:

-

1 (Greening): Significant decline in fossil fuel use across the majority of indicators

-

0 (Not Greening): Stagnation or increase in one or more key fuel types.

The Random Forest classifier was selected due to its robustness to small sample sizes, ability to model non-linear relationships, and interpretability via feature importance metrics. Alternative classifiers, including logistic regression and single decision trees, were tested during model development, but produced lower stability in classification accuracy, particularly under class imbalance. Random Forest provided the best F1-score for the “greening” class and exhibited lower variance across validation runs.

Given the small number of samples (n = 9 countries), we employed Leave-One-Out Cross-Validation (LOOCV) for model evaluation, which helps avoid overfitting and ensures the full utilisation of the dataset. The classifier was implemented using scikit-learn with the following parameters: 100 estimators (trees), Gini impurity as the split criterion, and automatic maximum depth selection. Feature importance scores were extracted using both Gini-based and permutation-based methods.

While the input features are interpretable in isolation, their joint interactions—particularly across multiple fuels and time windows—are non-trivial. For instance, a country may reduce its gas consumption, but increase its oil consumption, or vice versa, creating complex, multidimensional signals. The machine learning classifier helps capture these interaction effects and provides an empirical, reproducible rule for transition labelling that goes beyond visual interpretation.

The model was trained separately on the historical and recent time windows to identify both long-term transition status and emerging trends. Performance was evaluated using accuracy, precision, recall, and F1-score, with special attention given to the interpretation of the confusion matrix due to class imbalance. To assess model transparency and interpretability, feature importance scores were extracted using both Gini impurity and permutation-based metrics.

3.5. Temporal Classification Comparison and Label Reversal Detection

A key analytical innovation of this study is the identification of classification label reversals—cases where a country’s predicted status changed between the historical and recent models. These reversals are used as empirical markers of transition volatility, signalling either acceleration (positive shift) or deceleration (negative shift) in energy behaviour.

Additionally, a slope differential metric was calculated to quantify the rate of change over time. This metric is visualised across countries and fuel types to detect latent acceleration or rebound effects and to supplement binary classification with a richer picture of behavioural dynamics.

4. Results

4.1. Classification Performance of the Green Transition Model

To evaluate how much EU countries are undergoing structural decarbonisation, a Random Forest classifier was trained using time-series features that represent fossil fuel consumption trends. The model achieved an overall accuracy of 66.7% in predicting transition status based on historical patterns, with an F1-score of 0.80 for countries displaying “green” behaviour (

Table 1. Classification Performance.

Metric	Class 0 (Not Greening)	Class 1 (Greening)
Precision	0.00	0.86
Recall	0.00	0.75
F1-score	0.00	0.80
Accuracy	66.7% (6/9 countries)

).

However, the classification performance for the “not greening” class was poor, primarily due to class imbalance in the dataset—most countries showed at least some reduction, which skewed the distribution. Precision and recall for the “not greening” class were both 0.00, suggesting a need for more balanced training data or resampling strategies.

Figure 1 illustrates the confusion matrix of the historical-trend model, highlighting the imbalance between classes and the model’s difficulty in accurately identifying non-transitioning countries.

As visualised in Figure 1, the model performs well on ‘greening’ cases, but fails to capture reversal or stagnation. This shortcoming prompted the development of a second model, which focuses on recent trends, as discussed in the following subsection.

4.2. Classification Based on Recent Fuel Trends

To capture more immediate signals of policy impact, a second classification model was trained using only the most recent five years of fossil fuel consumption data. This short-term model significantly outperformed the historical one, achieving an overall accuracy of 88.9%, with a near-perfect F1-score of 0.93 for the “greening” class and 0.67 for the “not greening” class.

This suggests that recent behavioural shifts, likely influenced by post-2020 policy actions, pandemic-related energy dynamics, and REPowerEU directives, are more predictive of transition status than long-term historical inertia. Notably, this model was better able to detect countries not undergoing transition, indicating improved class balance in short-term trends.

Table 2. Metrics for Recent Model.

Metric	Class 0 (Not Greening)	Class 1 (Greening)
Precision	1.00	0.88
Recall	0.50	1.00
F1-score	0.67	0.93
Accuracy	88.9% (8/9 countries)	—

summarises the precision, recall, and F1-scores for both classes in the recent trend model, highlighting its improved ability to detect non-transitioning countries.

Figure 2 presents the confusion matrix for the recent-trend classification model, highlighting its ability to identify both transitioning and non-transitioning countries accurately.

As visualised in Figure 2, the recent-trend model performs with strong class balance and precision, supporting its use as a tool for evaluating near-term policy effectiveness in the post-2020 context.

4.3. Feature Importance and Model Interpretability

To understand what drives the classification of countries as either “greening” or “not greening,” feature importance was analysed using both Gini-based impurity and permutation importance methods. In both historical and recent models, the most influential features were related to liquid fuel consumption specifically: Liquid_slope—the rate of change in liquid fuel use, Liquid_delta—the net absolute change over the period (Figure 3).

These two features accounted for over 40% of total model importance in both timeframes, indicating that liquid fuels serve as the clearest empirical signal of green transition behaviour. This aligns with the political visibility and economic centrality of oil consumption in the EU’s energy mix, particularly in the transport and industrial sectors.

Gas and solid fuel indicators made a marginal contribution to the historical model. Still, they gained slight importance in the recent model, particularly Gas_slope_recent, reflecting more recent volatility in gas consumption due to geopolitical and market disruptions.

Among all predictors, only changes in liquid fuel usage had a measurable impact on classification performance. This suggests that policy efforts aimed at reducing oil dependence are the most empirically detectable signals of transition, reinforcing the political importance of transport and industrial decarbonisation pathways.

4.4. Temporal Dynamics and Label Reversals

Beyond binary classification, this study examines the temporal consistency of countries’ transition statuses by comparing predictions based on long-term historical trends with those derived from the most recent five-year window. This dual-window approach enables the identification of label reversals, where countries switch from “not greening” to “greening” or vice versa, revealing non-linear and dynamic decarbonisation paths.

Several EU Member States experienced such reversals. For example, Austria, Belgium, and Spain were initially classified as “not greening” based on historical data; however, they have shown substantial recent declines in fossil fuel consumption, leading to a “greening” classification in the short-term model. These upward reversals suggest that recent policy actions or post-pandemic structural shifts have begun to have a measurable impact on energy consumption patterns. In contrast, countries like Hungary and the Czech Republic displayed the opposite pattern: despite historical downward trends, their recent fossil fuel use either stagnated or rebounded. These downward reversals may signal transition fatigue, an economic recovery-driven rebound in emissions, or delayed policy implementation. Figure 4 below presents these label dynamics:

As shown in Figure 4, multiple EU Member States have changed their classification when comparing historical and recent fossil fuel trends. Countries such as Austria, Cyprus, and Spain moved into the “greening” category, suggesting more ambitious or recently intensified efforts to reduce fossil fuel use. In contrast, Hungary, Poland, and Malta shifted toward the “reversing” class, indicating potential backsliding or stagnation. These reclassifications underscore the limitations of relying solely on long-term trends and emphasise the importance of incorporating recent dynamics when evaluating transition progress.

To further quantify these behavioural shifts, we introduce the slope differential metric:

$$\Delta slope={\mathrm{slope}}_{\mathrm{recent}}-{\mathrm{slope}}_{\mathrm{historical}}$$

(1)

where Δslope denotes the recent acceleration or deceleration in fossil fuel consumption.

This metric (1) measures the acceleration or deceleration of fossil fuel consumption for each country and fuel type (gas, liquid, solid). Negative values indicate a faster decline in recent years (i.e., policy acceleration), while positive values suggest a slowing of progress or even a trend reversal. The results are visualised in Figure 5, a heatmap of slope changes:

This temporal analysis reveals that transition trajectories are far from static. While some countries have gained momentum, others risk losing ground, underscoring the need for continuous monitoring and adaptable policy responses.

To triangulate the slope differential findings and provide a clearer view of country-level momentum, two additional visual diagnostics were created. Figure 6 offers a direct comparison of the total fossil fuel trend slopes for each country across two periods: historical and recent. This chart highlights the direction and size of overall change. Countries such as Denmark (DNK), Spain (ESP), and Estonia (EST) show more negative slopes in the recent period, suggesting their transitions have not only continued, but become stronger. Conversely, countries like Slovakia (SVK) and Portugal (PRT) display flatter or less harmful recent slopes, indicating a slowdown or partial reversal in fossil fuel reduction efforts.

To capture fuel-specific dynamics, Figure 7 disaggregates the ∆Slope values by gas, liquid, and solid fuels. This offers insight into which fuel types are driving transition behaviour in each country.

Together, these visuals confirm that energy transitions in the EU are heterogeneous, fuel-dependent, and time-sensitive. These diagnostics also provide the empirical grounding for the country-level profiles that follow.

4.5. Country-Level Cases

To complement the aggregate classification and trend analyses, this section presents illustrative case studies of selected EU Member States. These country-level profiles are based on observed differences between historical and recent trends in fossil fuel consumption, with a focus on changes in slope and classification outcomes. The selected cases represent diverse trajectories: late-onset acceleration, structural deepening, and signs of reversal.

Austria represents a case of delayed but measurable progress. Historically, the country has exhibited minimal reductions in fossil fuel consumption, resulting in a “not greening” classification. However, over the recent five-year period, Austria has demonstrated moderate improvements in slope, particularly in the use of liquid fuel. This change resulted in a classification reversal, suggesting the emergence of effective policy mechanisms post-2020, possibly in the transport or heating sectors.

Spain exhibits one of the most significant improvements in recent slope trends, primarily driven by reductions in liquid and solid fuel consumption. Although its historical trajectory was moderate, recent dynamics indicate an intensified phase of transition. This can be attributed to a combination of EU-supported renewable energy expansion, national electric mobility incentives, and declining demand in the fossil fuel-based power sector.

France presents a rare case of multidimensional acceleration, with substantial recent declines observed across all three fuel types: gas, liquid, and solid. The steepening of these slopes indicates a structural rather than cyclical decarbonisation pattern. An integrated policy mix, including carbon taxation, energy efficiency programmes, and fossil-to-renewable substitution in district heating and industry, likely supports this trend.

In contrast, Hungary exhibits signs of stagnation or reversal. Despite a historically declining trajectory, recent slopes suggest flattening or slight rebounds in fossil fuel consumption. This has led to a downgraded classification in the recent window. The underlying causes may include economic recovery effects following the COVID-19 pandemic, short-term prioritisation of energy security, or a lack of implementation capacity at the national level.

These cases collectively highlight the diversity and complexity of transition pathways within the EU. While some countries demonstrate a strengthening of decarbonisation efforts, others reveal the fragility of early progress. The observed temporal volatility emphasises the need for sustained adaptive and sector-specific interventions to secure long-term transition outcomes.

5. Discussion

This study aimed to investigate the extent to which the European Union’s climate rhetoric—enshrined in flagship initiatives such as the European Green Deal and Fit for 55—is reflected in the empirical record of national-level fossil fuel consumption. By combining long-term and recent behavioural trend analysis with machine learning classification, the findings offer a nuanced picture: while rhetorical ambition exists at the supranational level, its translation into material transition outcomes is partial, asymmetric, and temporally inconsistent.

The results reveal that a significant share of Member States have demonstrated recent progress in decarbonization, particularly in liquid fuel consumption, which has emerged as the most sensitive and policy-responsive indicator. This trend aligns with the increasing EU-level regulation in the transport sector, including vehicle electrification mandates and fuel taxation. It may signal that rhetorical commitments are beginning to influence real-world energy systems. In countries like France and Denmark, where fossil fuel decline was both deep and cross-sectoral, the evidence supports the notion that rhetorical intent and material transformation can be mutually reinforcing.

However, the inclusion of slope differential metrics and dual-window classification introduces a more complex narrative. Many countries classified as “greening” within the historical window exhibited stagnation or even reversal in recent years. These label reversals, observed in countries such as Hungary and the Czech Republic, raise important questions about the durability of transition trajectories and the extent of policy institutionalisation. Short-term improvements may reflect temporary demand shocks or cyclical effects rather than structural change. Conversely, recent gains in late-shifting countries like Austria and Spain suggest that the transition is not necessarily linear, but can be reactivated under the right political and economic conditions.

These findings reinforce the idea that rhetoric and reality are misaligned, not categorically, but conditionally. The EU’s strategic climate vision sets clear normative expectations, but the operationalisation of that vision varies widely across Member States. Multilevel governance, divergent national energy mixes, and the uneven absorption of EU funding and regulatory instruments all contribute to this disparity. In particular, the discrepancy between long-term and recent trends underscores the vulnerability of transition processes to policy fatigue, political turnover, or energy price volatility, especially in the wake of overlapping crises such as the COVID-19 pandemic and the war in Ukraine.

From a methodological perspective, this study demonstrates the importance of combining classification modelling with time-series diagnostics to understand transition dynamics. By implementing a dual-window approach and the slope differential metric, the analysis goes beyond static “green” versus “not green” categories to uncover momentum, acceleration, and reversibility—dimensions that are essential for policy assessment, but are often missed in headline indicators.

6. Policy Implications

Our findings have significant implications for both EU-level governance and national policymaking.

1. The need for differentiated transition monitoring: Binary labels and aggregate targets obscure meaningful differences in transition pace and sectoral composition. EU oversight mechanisms should incorporate fuel-specific, time-sensitive indicators to detect where momentum is building or eroding.

2. Addressing policy volatility: The presence of label reversals in multiple countries signals the fragility of transition pathways. This underscores the need to institutionalise climate action across electoral cycles, reduce regulatory uncertainty, and align fiscal instruments (e.g., green subsidies, taxation) with long-term goals.

3. Reinforcing the link between rhetoric and enforcement: Political commitments must be accompanied by robust implementation architectures. Strengthening mechanisms such as the NECP review process and climate conditionalities in EU funding could help close the gap between aspiration and execution.

4. Early-warning diagnostics for policy adaptation: Tools like slope differential tracking and machine learning classification can act as early warning systems, flagging deceleration or reversal before formal compliance gaps appear. This empowers policymakers to intervene proactively rather than retroactively.

In summary, the European Union’s climate narrative is not entirely disconnected from material outcomes; however, the relationship is conditional, uneven, and sensitive to time. Ensuring that rhetoric continues to align with and reinforce empirical trends will require sustained political commitment, rigorous monitoring, and adaptive governance mechanisms that can navigate the volatility of transitions.

7. Limitations and Future Research

This study focused on empirically classifying national patterns of fossil fuel consumption and their alignment with climate discourse. While our results identify patterns of transition volatility and reversal, we do not claim to explain the underlying causes of these patterns. Future research could investigate the impact of structural national differences—such as economic development, industrial composition, and governance models—on shaping energy trajectories across the EU. Integrating such socioeconomic factors would allow for a more explanatory framework, complementing the trend-based diagnostics presented here.

Additionally, future studies could examine the role of intra-EU migration and transnational population mobility, which may influence national energy consumption patterns and complicate the attribution of decarbonisation trends to domestic policy effectiveness.

8. Conclusions

This study provides a data-driven evaluation of how well the European Union’s climate goals align with the actual energy transition behaviours of its Member States. The findings suggest a cautiously optimistic but fundamentally uneven relationship between words and actions. On the one hand, several countries—most notably Germany, Denmark, and Austria—have shown clear declines in fossil fuel use, particularly in liquid fuels. These patterns, identified through machine learning classification, imply that EU-level policies may be influencing national behaviours in a tangible way.

However, the analysis of slope acceleration metrics and label reversals uncovers a more intricate and fragmented picture. Some Member States, despite public promises for green transition, exhibit stagnation or even reversals in certain fuel types. This indicates that transition pathways are neither straightforward nor all moving forward at the same pace; instead, they are influenced by short-term shocks, reactive policy adjustments, and uneven implementation. Countries switching between “green” and “not green” labels further highlight that energy transition is a complex process, not a simple binary status, and requires ongoing empirical review. Phrases like “Fit for 55” and the “Green Deal” set important normative goals, but these must be continually checked against actual behaviours and structures.

In short, while some EU Member States are indeed acting accordingly, others remain caught between commitment and real change. This underscores the importance of continuous, multi-faceted monitoring of decarbonisation efforts and the need to be cautious about equating aspirational stories with real progress.