Does Financial System Stability Foster Sustainability? Evidence from European Union Countries

1. Introduction

Sustainable economic development requires sustained investment, firms able to adapt to economic change, and employment opportunities. The financial system supports these processes by financing firms and allocating resources across activities. Disruptions to financial intermediation can restrict access to credit, delay investment projects and increase the costs of adjustment. The European Central Bank defines financial stability as the system's ability to withstand shocks and avoid major disruptions to financial intermediation and real economic activity [1]. This resilience matters for a sustainable economy, but its environmental and social contribution also depends on the activities that finance supports.

Financial resilience and the contribution of finance to sustainability address different questions. Well-capitalised banks may continue to finance high-emission activities, while calm financial markets do not necessarily support investment in renewable energy. At the same time, the transition to a lower-carbon economy requires investment in new infrastructure and technologies and may alter firms' risk profiles and financing needs. As Campiglio explains, credit market imperfections and lending decisions can constrain low-carbon investment even when carbon pricing provides an incentive to invest [2]. Whether financial conditions support sustainability therefore depends both on the continuity of financing and on where financial resources are directed.

European Union policy brings these questions together. The European Commission's strategy for financing the transition to a sustainable economy emphasises economic transformation, inclusiveness and financial resilience [3]. Pursuing these objectives together does not guarantee equally favourable outcomes across all dimensions. Sustainable finance concerns the allocation of resources towards environmental and social objectives, whereas financial stability concerns the system's ability to function under adverse conditions [4]. Examining environmental and social outcomes separately helps establish how these related policy aims align in practice.

Previous studies do not show a uniform relationship between financial stability and sustainability. In EU countries, Şahin and Arzova [5] report different associations with green growth for banking stability and access to credit. Ozili and Iorember [6] find that associations vary across sustainability indicators and economic contexts. Hoang [7] reports heterogeneous links between financial stress and environmental outcomes, while Javed et al. [8] examine how macro-financial stability shapes the finance–green-growth relationship in developing economies. These studies use different financial measures, samples and outcomes, making it difficult to generalise their findings to environmental and social changes in the EU. Together, these studies raise a more specific question for the EU: do distinct aspects of financial stability accompany similar changes in environmental and social outcomes?

Three issues motivate that question. First, the stability indices and risk measures used in the previous studies capture different financial conditions. Their results leave open the question of whether bank capital adequacy, loan portfolio quality, and market stress have similar relationships with sustainability. Second, studies centred on green growth or environmental indicators give less insight into whether environmental improvements coincide with social progress. Third, the choice of environmental measure matters: a higher share of renewable energy, more generating capacity, and lower emissions describe different aspects of the transition. Examining these issues within an EU framework can reveal which relationships recur across measures and which are specific to a particular outcome.

This study aims to assess the associations between different dimensions of financial stability and changes in environmental and social sustainability indicators across European Union countries. Financial stability is examined through three measures: the Tier 1 capital ratio, the non-performing loan ratio, and the Country-Level Index of Financial Stress (CLIFS). CLIFS and Tier 1 define the two primary analyses; the non-performing loan ratio is examined in a separate banking comparison and exploratory common-sample models. Environmental sustainability is assessed using renewable energy expansion and changes in per-capita greenhouse gas emissions. At the same time, the social dimension is represented by changes in the employment rate among people aged 20–64. A supplementary composite sustainability index covering environmental, social, and economic dimensions is also examined. This approach enables assessment not only of the associations between individual financial stability measures and selected sustainability outcomes, but also of whether these associations are consistent across sustainability dimensions.

The study uses two unbalanced EU panels: a financial market-stress panel for 2010–2024 and a bank-capitalisation panel for 2015–2024. Models with country and year fixed effects relate financial conditions to annual outcome changes. Subsequent exploratory analyses compare financial measures on common country-years, add controls relevant to each outcome, and examine a composite sustainability index and nonlinear relationships. Section 3 explains the samples, timing and measures; the appendices provide the technical details.

The study provides evidence on how distinct aspects of financial stability relate to environmental outcomes and employment. It examines capital adequacy, loan quality and market stress separately, and places environmental outcomes alongside employment. The common-sample comparisons show how the findings change with outcome-specific controls, while the supporting index reveals which components account for an aggregate association. This approach makes it possible to assess how far a broad claim about financial stability and sustainability is supported by the individual relationships.

Higher financial market stress is associated with smaller annual increases in the employment rate in the primary and exploratory common-sample analyses. However, the primary association is not statistically significant after adjustment for the three outcomes, and the exploratory result depends on how many tests are included in the adjustment. The environmental estimates do not establish a consistent favourable association with lower stress or higher bank capitalisation. The composite index result largely reflects its employment component. Overall, the evidence is more specific to employment than to sustainability outcomes as a whole.

The remainder of the article reviews the theoretical and empirical literature, describes the data and methods, presents the results, discusses their research and policy implications, and concludes.

2. Financial Stability and Sustainability: Mechanisms and Evidence

2.1. Financing Continuity and Low-Carbon Investment

Financing continuity provides one reason to expect a relationship between financial stability and sustainability. Renewable energy infrastructure and production changes often require large initial investments, with returns and financial obligations extending over many years. Disruptions to credit can delay these projects or threaten their completion. Resilient intermediaries may be better able to absorb losses and maintain lending when economic conditions deteriorate. The environmental contribution, however, depends on financing projects that reduce emissions, expand renewable energy, or introduce cleaner technologies. Campiglio (2016) explains how credit market imperfections, lending decisions and project risk–return characteristics can constrain low-carbon investment, and discusses the potential role of monetary and macroprudential policies in addressing these barriers [2].

The relationship between financial system conditions and environmental sustainability is not straightforward, as access to finance does not necessarily ensure that investment is directed towards environmentally sustainable activities. Mensah and Näsström (2022), using a broad international sample, examine the relationship between banking sector performance and renewable energy consumption and find that the results differ across country groups [9]. Beka and Abazi-Alili (2024), analysing 38 OECD countries and five Western Balkan countries over 2010–2022, similarly identify differences in the associations between banking sector performance indicators and renewable energy consumption across the two country groups [10]. OECD countries report a statistically significant positive association between financial stability, measured by the banking system Z-score, and renewable energy consumption. In the Western Balkan countries, the association with the Z-score was negative but not statistically significant at the 5% level. In contrast, a statistically significant positive association was reported for the non-performing loan ratio, which the authors use as an asset-quality indicator. Since a higher non-performing loan ratio indicates poorer loan portfolio quality, this result should not be interpreted as evidence that improvements in asset quality are positively associated with renewable energy consumption. Overall, these findings suggest that the relationships between banking sector conditions and environmental outcomes may vary across economic and financial environments.

Renewable energy consumption captures one part of this relationship. Financial systems can finance clean energy while also supporting activities that raise total energy demand and emissions. Examining renewable energy expansion alongside greenhouse gas emissions, therefore, provides a fuller view of the environmental changes associated with financial conditions.

Financial resilience and access to investment finance can also move differently. A higher capital adequacy ratio may result from more regulatory capital, fewer risk-weighted assets, or both. Banks can therefore improve the ratio while restricting lending. Better loan quality may strengthen their capacity to finance new projects, but the destination of that finance remains a separate question. Capital and loan-quality ratios describe resilience and credit risk; they do not directly measure lending to environmentally sustainable activities.

The distinction between financial stability and access to finance is also illustrated by Şahin and Arzova (2026), who examined the relationships between political, macroeconomic, and financial stability and green economic growth using data from 27 EU countries over 2012–2021 [5]. Their findings indicate that the association between banking stability, measured by the Z-score, and green economic growth was not statistically significant in the full sample. Domestic credit to the private sector showed a positive association, but was statistically significant only at the 10% level. Additional subgroup analyses revealed a weak positive association between credit provision and green economic growth in the more economically developed EU countries. In contrast, the corresponding association was not statistically significant in Central and Eastern European countries. These findings suggest that financial system resilience and access to finance may exhibit different relationships with environmental outcomes and that these relationships may vary with countries’ economic and institutional characteristics. This provides a rationale for separately examining the associations between various financial stability indicators and selected environmental sustainability outcomes, including renewable energy expansion and per-person greenhouse gas emissions.

2.2. Capitalisation, Asset Quality and Financial Market Stress

Assessing the relationship between financial stability and sustainability requires consideration of different aspects of financial system conditions, as bank capital adequacy, asset quality, and financial market stress reflect distinct mechanisms of financial resilience and risk. The Tier 1 capital ratio, defined as Tier 1 capital relative to risk-weighted assets, indicates banks’ capacity to absorb potential losses. The non-performing loan (NPL) ratio reflects the quality of the loan portfolio and the level of credit risk. Meanwhile, the Country-Level Index of Financial Stress (CLIFS) captures financial market stress associated with asset price fluctuations, increased uncertainty, and co-movements across financial market segments. Duprey et al. (2017) developed a methodology for identifying episodes of systemic financial stress. They demonstrated that periods of elevated financial market stress may be associated with substantial declines in real economic activity. They also found that, in most identified systemic stress episodes, financial market stress preceded real economic stress [11]. These findings support the relevance of financial market stress measures when examining the relationship between financial stability and economic and sustainability outcomes.

These financial dimensions suggest different pathways to sustainability outcomes. Capital adequacy and better loan quality may help banks sustain lending through difficult periods. Market stress may accompany tighter financing conditions, postponed investment and more cautious hiring, affecting both renewable projects and employment. At the same time, an economic slowdown can reduce energy consumption and emissions. The same deterioration in financial conditions could therefore coincide with weaker employment and lower emissions, making separate outcome comparisons essential.

The complexity of the relationship between financial stability and sustainability is further illustrated by Ozili and Iorember (2024), who analysed data from 26 countries over 2011–2018 using the system generalised method of moments (system GMM) [6]. They found that the relationship between financial stability and sustainable development depends on the sustainability measure employed and countries’ economic characteristics. Their analysis of a composite sustainable development index identified a statistically significant association with financial stability, with a negative relationship reported for Asian countries. Analyses of individual Sustainable Development Goals (SDGs) also revealed heterogeneous findings. Notably, financial stability was positively associated with SDG 3 (Good Health and Well-Being) and SDG 7 (Affordable and Clean Energy) in countries with higher banking-sector capital buffers. These findings suggest that bank capital buffers may condition the relationship between financial stability and specific dimensions of sustainable development, and that the nature of the association depends on the sustainability outcome considered.

The role of financial stability in green economic growth was further investigated by Javed et al. (2026), who examined 34 developing economies from 1991 to 2022 [8]. Their study used a financial risk rating covering external financial vulnerabilities to examine whether macro-financial stability conditions the relationship between financial development and environmental outcomes. The authors reported that greater macro-financial stability strengthened the positive associations between financial institution and financial market development and green economic growth. Their interaction analyses also indicated more favourable associations with renewable energy consumption and lower carbon dioxide emissions intensity. These findings suggest that financial stability may be relevant not only as an independent characteristic of the financial system but also as a condition shaping the relationship between financial development and environmental outcomes.

The complexity of the relationship between financial market stress and environmental sustainability is also highlighted by Hoang (2026), who analysed 36 countries over 1991–2023 using panel quantile regression [7]. The results indicate that the associations between financial stress, renewable energy consumption, and carbon dioxide emissions vary across conditional quantiles of the outcome distributions. Both positive and negative associations were identified, suggesting that increases in financial stress are not uniformly associated with improved or deteriorating environmental outcomes. The study also reports adverse associations between geopolitical risk and the energy transition and environmental conditions. These findings complement previous research by indicating that the relationships between financial instability and environmental outcomes may depend on economic conditions and the characteristics of the indicators examined.

The literature thus provides reasons to distinguish between financial measures and outcomes. Bank capital ratios, loan quality, composite stability measures and macro-financial risk ratings describe different conditions. Studying capital adequacy, asset quality and market stress separately in EU countries allows us to compare their relationships with selected environmental and social changes without assuming that they represent equivalent improvements in stability.

2.3. Renewable Energy, Emissions and Employment

Improvements in environmental outcomes do not necessarily coincide with improvements in social outcomes. Renewable energy can expand through new infrastructure, technological improvements and public support. Emissions can decline because production becomes cleaner, energy use becomes more efficient, or economic activity contracts. Employment encompasses participation in economic activity and access to labour market opportunities. During periods of financial stress, a slowdown may therefore lead to both lower emissions and weaker employment. Assessing those outcomes separately distinguishes a cleaner economy from an economy that emits less because it produces less.

Renewable energy development can also be assessed using different indicators that capture distinct aspects of the energy transition. Tu et al. (2022), examining the determinants of renewable energy development in EU countries, highlight the importance of economic, technological, and institutional factors [12]. Valdés Lucas et al. (2016), investigating the relationship between renewable energy deployment and energy security in the EU, emphasise their interdependence and the relevance of energy system structure and reliance on external energy resources [13]. These studies suggest that renewable energy development depends on a range of economic and structural conditions, which may lead to heterogeneous associations with financial stability indicators. Moreover, an increase in the renewable share of energy consumption does not necessarily imply a proportional expansion of new generation capacity, as the change in the share also depends on total energy consumption. By contrast, the net change in installed wind capacity directly captures the expansion or contraction of physical generation capacity. Still, it does not measure actual electricity generation or the monetary value of investment. These indicators therefore provide complementary rather than equivalent information about the energy transition.

The STIRPAT framework developed by York et al. (2003) provides a theoretical basis for analysing changes in GHG emissions by relating environmental impacts to population, affluence, and technology [14]. This perspective helps explain why emissions are influenced not only by the energy transition but also by the scale of economic activity and production technologies. The European Environment Agency identifies several interrelated drivers of declining GHG emissions in the EU, including lower energy-production carbon intensity, improvements in energy efficiency, changes in the structure of economic activity, and climate policy measures [15]. At the same time, expanding economic activity may increase energy demand and emissions, making the relationship between financial stability and GHG emissions ambiguous. A financial system capable of supporting investment may facilitate the adoption of less carbon-intensive technologies while also financing broader economic expansion. Because these mechanisms may operate in opposing directions, the relationship between financial stability and changes in GHG emissions remains an empirical question.

Within the social dimension of sustainability, the employment rate captures an important aspect of economic participation. However, changes in employment depend not only on financing conditions but also on the business cycle, workforce qualifications, and demographic structure. Consolo et al. (2026), examining labour force developments in the euro area, highlight the roles of population ageing, educational attainment, and migration in shaping labour supply and labour force participation [16]. An, Bluedorn, and Ciminelli (2021), investigating Okun’s law across economies and worker groups, demonstrate that the cyclical sensitivity of unemployment varies with economic development and demographic characteristics [17]. Although these studies focus on labour force participation and unemployment rather than the employment rate used in the present study, they provide a basis for considering how economic and structural conditions may shape the relationship between financial market stress and employment. Increasing financial stress may restrict firms’ ability to finance operations and investment, encourage more cautious hiring decisions, and consequently be associated with weaker employment outcomes. However, the strength of these associations may vary according to labour market structure and broader economic conditions.

These differences form the basis for the empirical comparison. Renewable expansion, emissions and employment each require measures and controls suited to the outcome being studied. An index combining environmental, social and economic indicators can then summarise their joint movement. Reading that index alongside its components shows whether an aggregate association reflects broad progress or is concentrated in one dimension.

3. Data and Empirical Approach

3.1. Country Panels and Measures

The unit of observation is a country in an outcome year. The primary financial market-stress panel contains 371 country-year observations from 26 current EU members over 2010–2024. The primary bank-capitalisation panel contains 258 observations covering the current EU27 over 2015–2024. Estonia is excluded from analyses requiring CLIFS, including the stress and common-sample analyses, but remains in the primary capital analysis. Data availability and the exclusion of employment changes crossing reported statistical breaks produce unbalanced panels. The country group follows current EU membership; some countries were not members throughout the study period. Appendix A explains sample construction.

Financial stress is measured using the average monthly CLIFS value from the year preceding the outcome year. The annual average is calculated only when all 12 monthly observations are available. The Tier 1 and non-performing loan ratios use the preceding year-end observation from ECB consolidated banking data. A higher Tier 1 capital ratio indicates stronger bank capital adequacy, whereas higher NPL and CLIFS values indicate poorer asset quality and greater financial market stress, respectively. A five-percentage-point increase is the reported contrast for each banking ratio; a 0.1-index-unit increase is the contrast for CLIFS. These scales support readable interpretation within each measure. They are not a claim that the contrasts represent equivalent changes in risk. Table 1 summarises the financial measures, their interpretation and the corresponding study coverage.

The primary financial market-stress analysis examines annual changes in the share of renewable energy in gross final energy consumption, territorial greenhouse gas emissions per capita, and the employment rate among individuals aged 20–64. Renewable and employment changes are measured in percentage points. Changes in greenhouse gas emissions are measured in tonnes of CO₂ equivalent per capita, excluding emissions and removals from land use, land-use change and forestry (LULUCF), as well as separately reported emissions from international aviation, maritime transport and other memorandum items. The primary Tier 1 analysis uses annual net changes in installed wind power capacity, logarithmic changes in greenhouse gas emissions, and the same change in the employment rate. The wind outcome is the change in capacity between 31 December of consecutive years, divided by the population on 1 January of the outcome year, in millions. Its unit is megawatts per million residents. Emissions log changes are 100 times the logarithm of the ratio between successive years, expressed as log-per cent points.

The environmental measures provide complementary views of the transition. The stress analysis examines changes in the energy mix and absolute changes in per-person emissions. The capital analysis examines net additions in wind capacity and proportional changes in emissions. A decline in wind capacity is retained because it can reflect retirements or statistical revisions; it is a net stock change, not negative investment expenditure. All employment models use the rate among people aged 20–64. Appendix A documents the source series, flags and reporting parameters.

The sample rules preserve the distinction between economic changes and reporting changes. Employment changes crossing a reported break are excluded, while observations with remaining definition differences receive a separate sensitivity check. Missing financial observations are not replaced by zero, and missing controls are not interpolated. Differences between the primary panels therefore reflect both outcome definitions and available observations. The common-sample extension holds the observations fixed when comparing specifications within each outcome.

3.2. Within-Country Comparisons and Timing

The linear models compare annual changes within countries. Country fixed effects account for each country's typical annual-change pattern over the observed period; year fixed effects account for shifts shared across countries. The original controls are log real GDP per person, adjusted for purchasing power, and consumer-price inflation, both measured two years before the outcome. Each country-year receives equal weight. The comparison asks whether different levels of financial stress, bank capitalisation, and loan quality are associated with different annual changes in outcomes within countries.

The timing of financial stress is central to that interpretation. An annual outcome change compares two successive years, with the first year also being the CLIFS averaging period. Stress, therefore, precedes the endpoint but overlaps with the baseline of the change. The estimated relationship may reflect financial and economic conditions already developing together in that baseline year. Stricter timing checks move the financial exposure before the baseline and the controls earlier still, testing whether the pattern also appears with greater temporal separation. These lag choices help interpret the association, though the observational design remains vulnerable to confounding and feedback.

The uncertainty procedures allow observations from the same country to be related over time. In the capital models, a small-sample adjustment also accounts for the limited independent country-level information informing the estimate [18, 19]. This adjusts the standard errors, confidence intervals and tests, rather than the observed data or coefficient estimates. Within each primary analysis, Holm's adjustment considers the three outcome tests together. A separate bootstrap calculation checks how the results depend on the uncertainty procedure. This comparison matters particularly for capitalisation, where much of the informative variation comes from a few countries. Appendix B gives the covariance methods, degrees of freedom, interval definitions and groups of tests used in each adjustment.

3.3. Exploratory Comparisons and Aggregation

We distinguish the two primary analyses from later exploratory comparisons. The market-stress analysis plan was fixed locally before its estimation. The bank-capital analysis was developed after the market-stress results were known, and the tailored-control models, composite index and nonlinear analyses followed both primary analyses. This sequence explains which questions guided each analysis and which arose from the preceding findings; it does not constitute external preregistration.

The subsequent comparisons use common recent country-years for all three financial measures and incorporate additional controls selected to reflect economic and structural factors relevant to each outcome. The wind capacity specifications also account for economic growth, energy import dependence, and existing installed wind capacity. The emissions specifications include economic growth, manufacturing share and energy intensity, while the employment specifications control for economic growth, tertiary educational attainment and the share of older individuals in the working-age population. These controls are dated two years before the outcome. The matched original-control models use the same country-years as their tailored counterparts, allowing a clearer comparison of specifications. Appendix D explains the controls and retained source sensitivities.

The models examining changes in installed wind power capacity and employment rates are each estimated using 246 country-year observations covering 26 countries. The emissions specifications use 238 country-year observations across 25 countries because eight Bulgarian manufacturing observations are unavailable and are not imputed. The nine financial coefficients from the tailored-control models are treated as a single family of tests for multiple-testing adjustment. A broader multiple-testing adjustment covers 57 designated exploratory tests, including financial coefficients from additional model specifications and tests of nonlinear relations. Reporting both adjustments makes clear how conclusions change when a broader set of hypotheses is considered. Differences between coefficients are descriptive unless a direct statistical contrast is provided. Neither the common sample nor its tailored controls displace the original primary results.

Finally, a supporting index combines the renewable energy share, favourably oriented emissions per person, employment, and real income per person. It follows the principle that normalisation, weighting and sensitivity should be transparent [20]. Each component is standardised against the fixed EU27 distribution in 2014. The environmental, social, and economic pillars receive equal weight, with the two environmental components each accounting for half of the pillar's weight. The main index regressions use 246 eligible country-year observations. Annual-change and level models form an exploratory family of six tests, with a broader adjustment reported as well. Appendix E provides construction and alternatives; Appendix F reports the nonlinear comparisons. These exploratory comparisons supplement the primary results.

4. Results

4.1. Financial Market Stress and Employment Progress

Higher financial market stress accompanies weaker annual employment improvement in the primary EU26 panel. A 0.1-point higher prior-year CLIFS is associated with a 0.305-percentage-point smaller change in the employment rate among people aged 20–64. This means a smaller increase, or a larger decline, in the share employed. The marginal 95% interval runs from −0.574 to −0.037 percentage points. After considering the three outcomes together, however, the Holm-adjusted p-value is 0.083; the separately adjusted bootstrap p-value is 0.206. Neither procedure rejects the null hypothesis at the 5% level.

Figure 1 places this employment pattern beside the environmental results from both primary analyses. The point estimates suggest that financial conditions carry different information about labour-market adjustment and environmental progress. The environmental intervals allow associations in either direction, while the negative stress–employment estimate remains uncertain after the three-outcome adjustment.

The negative employment estimate is not driven by a single country. Removing each country in turn leaves all 26 coefficients negative, ranging from approximately −0.343 to −0.183 percentage points. The direction thus persists across samples with different combinations of national banking systems and labour markets.

The timing check gives a different picture. When stress is measured before the baseline year of the employment change, the estimate is positive and imprecise: 0.0666 percentage points per 0.1-point increase in CLIFS (95% CI: −0.3021 to 0.4353; p = 0.713). A model containing both stress lags gives a negative but imprecise sum of −0.1229 percentage points (95% CI: −0.5319 to 0.2862; p = 0.542). The recurring negative pattern therefore concerns stress during the baseline year and the annual change that follows, when financial and labour-market conditions may already be evolving together. Appendix C reports the timing and country-deletion checks.

4.2. Environmental Progress and Bank Capitalisation

The environmental estimates allow for both favourable and unfavourable relationships. In the primary financial market-stress analysis, a 0.1-unit increase in CLIFS is associated with a 0.060-percentage-point decrease in the renewable energy share (95% interval: −0.307 to 0.188). The emissions estimate is 0.027 tonnes of CO₂ equivalent per person, with an interval from −0.065 to 0.118. Both point estimates suggest less favourable environmental changes under greater stress, but each Holm-adjusted p-value is 1.000. These estimates do not establish an environmental advantage associated with lower stress.

The bank-capitalisation analysis also produces mixed results, with no clear pattern emerging across the outcomes. A five-percentage-point higher Tier 1 ratio is associated with 15.66 MW less in annual net wind capacity expansion per million residents. Its primary 95% interval extends from 37.72 MW less to 6.41 MW more, with a three-test adjusted p-value of 0.354. The negative point estimate is substantively different from the expected financing-continuity benefit, but the primary uncertainty does not establish that higher capitalisation reduces wind expansion. For emissions, the coefficient is −1.061 log-per cent points, with an interval from −3.250 to 1.128. Employment is near zero at −0.031 percentage points, with an interval from −0.449 to 0.388.

For wind capacity, statistical support depends on the uncertainty method. The primary capital test does not reject at the 5% level, whereas the separately adjusted restricted wild-cluster bootstrap gives a p-value of 0.041. Both assess the same negative coefficient; the disagreement concerns its statistical significance. We use the primary analysis for the main conclusion and report the bootstrap result as evidence of sensitivity to the inference method. Figure 1 shows the primary intervals; Appendix C provides both sets of adjusted tests.

The distribution of capital variation helps explain this sensitivity. Estonia accounts for almost half of the Tier 1 variation remaining after other model factors are accounted for. Although the panel covers 27 countries, much of the information about the capital relationship comes from a single national experience. The primary uncertainty procedure accounts for that concentration and produces a wide wind interval. Appendix C reports the concentration and degrees-of-freedom measures.

Taken together, the environmental results show why the measures need to remain distinct. Renewable shares describe the energy mix, wind capacity records physical expansion, and emissions reflect both production and technology. Figure 1 uses a separate scale for each outcome. The clearer recurring pattern concerns employment, while understanding the environmental relationships will require data linking financing conditions more closely to the type and timing of transition activity.

4.3. Comparing Financial Dimensions with Outcome-Specific Controls

In the common recent sample, financial market stress continues to accompany weaker employment improvement, while capitalisation and asset quality do not exhibit a generally favourable relationship across outcomes. The tailored employment estimate is −0.226 percentage points per 0.1-point increase in CLIFS, with a marginal 95% interval from −0.347 to −0.105. Its adjusted p-value is 0.015 across the nine tailored coefficients, but 0.093 across the broader set of 57 designated tests. The exploratory estimate has the same negative sign as the primary estimate, although differences in samples, controls and testing groups prevent a direct comparison of statistical support. Its adjusted statistical significance also depends on which hypotheses are included in the multiple-testing adjustment.

Figure 2 compares original and tailored controls on identical country-years within each outcome. Holding the sample fixed shows how the estimates change when economic demand and outcome-specific circumstances are included in the comparison. For employment, the additional factors are growth, educational attainment, and the age composition of the working-age population. The estimated association between financial market stress and annual changes in the employment rate remains negative after accounting for the additional control variables. The analysis of installed wind power capacity also considers existing installed capacity and energy import dependency, while the greenhouse gas emissions analysis accounts for manufacturing share and energy intensity. These additional controls help interpret the results in the context of relevant economic and energy-sector characteristics, while recognising that the outcomes measure different aspects of sustainability.

Capitalisation and non-performing loans show no common favourable pattern across wind, emissions and employment. The signs and uncertainty differ by outcome. Their interpretation also differs: a higher Tier 1 ratio indicates more capital relative to risk exposure, whereas a higher NPL ratio indicates poorer loan quality. The separate estimates identify the stress–employment pattern as a specific relationship, rather than a result shared by all financial measures.

The employment direction persists after additional country deletions in the tailored model: all 26 omitted-country coefficients are negative, ranging from approximately −0.245 to −0.195 percentage points. A sensitivity analysis excluding flagged education and age-composition controls retains a negative estimate of about −0.286, but uses only 180 country-years and has a nominal p-value near 0.040. The smaller sample is a material part of that comparison. Source flags deserve attention, yet deleting flagged observations changes both the information and the coverage; it cannot be interpreted as a purely measurement correction applied to an otherwise identical panel. Appendix D reports the matched comparisons and relevant source sensitivities.

The exploratory comparisons add a specific finding to the primary analysis: the negative stress–employment pattern recurs after accounting for growth, education and age composition. Financial market stress is the measure most consistently linked to weaker employment progress in these models. The environmental and bank-ratio evidence remains less settled. The smaller emissions sample excludes Bulgaria, and changes in national policies, investment conditions, and other unmeasured factors may still shape the observed relationships.

4.4. What an Aggregate Sustainability Measure Adds

The supporting annual sustainability index reproduces the negative association with financial market stress mainly through employment. A 0.1 higher CLIFS is associated with a −0.0225-point annual index change, with a marginal 95% interval from −0.0356 to −0.0093. The six-test adjusted p-value is 0.019; the broader 57-test adjustment gives 0.176. Statistical support is weaker when all 57 exploratory questions are considered together. Within the focused index comparison, it gives a coherent summary of slower progress under greater financial market stress. The score uses fixed component standardisation and weights, so its units are weighted component score points rather than percentages or an official sustainable-development rating.

Figure 3 shows the annual index estimates alongside the point decomposition of financial market stress. The employment-based social pillar supplies approximately 57.3% of the negative CLIFS coefficient. A negative index association can be driven by weaker employment progress even if environmental relationships remain uncertain. Aggregation thus makes the outcome pattern easier to summarise while also illustrating how a broad label can hide its principal component.

The negative financial market stress sign persists across all 36 pillar-weight combinations and under income exclusion. Its nominal p-value nevertheless rises to approximately 0.159 in the weight comparisons. The sign is therefore less sensitive to these choices than the degree of precision. Capitalisation is more sample-sensitive: its annual-index coefficient is positive on its own 258 available country-years and negative on the 246 matched country-years, with neither estimate precise. No level result passes the six-test adjustment. Appendix E records these comparisons. The nonlinear checks in Appendix F provide no adjusted evidence for the specified curvature or threshold. Their small, inconsistent improvements when predicting omitted country-year observations do not change the substantive interpretation of the linear results.

5. Discussion

The EU evidence points to different relationships across financial and sustainability dimensions. The recurring negative stress–employment pattern stands out relative to the environmental and banking-ratio estimates. The two primary analyses cover different periods, samples and environmental measures, while loan quality enters chiefly through secondary and exploratory analyses. The contribution is therefore a comparison of specific relationships, with common-sample extensions helping to distinguish changes in specification from changes in coverage.

The most persistent directional pattern concerns financial market stress and employment. In the original panel, a 0.1-unit increase in CLIFS was associated with a 0.305-percentage-point smaller annual change in the employment rate. Although the marginal confidence interval excluded zero, neither the three-outcome Holm adjustment (p = 0.083) nor the separate bootstrap adjustment (p = 0.206) rejected at the 5% level. The exploratory estimate with employment-specific controls was also negative (−0.226 percentage points), passing the nine-test adjustment (p = 0.015) but not the broader 57-test adjustment (p = 0.093). When financial market stress was measured before the baseline year for the employment change, the estimate was positive and imprecise. Thus, the repeated negative estimates concern models in which measured stress overlaps the baseline year; they do not demonstrate a delayed employment response or show that reducing stress would increase employment.

The environmental findings provide no consistent evidence of a favourable association with greater financial stability. Estimates of renewable share and per-person emissions in the primary stress analysis were imprecise. In the primary capital analysis, a 5-percentage-point higher Tier 1 ratio was associated with 15.66 MW less in annual net wind capacity change per million residents. The primary adjusted test did not reject the null hypothesis (p = 0.354), whereas the separate bootstrap test did (p = 0.041); statistical support for the negative estimate therefore depends on the inference procedure. The capital–emissions estimate was also imprecise, and the exploratory NPL estimates did not show a consistent favourable pattern across wind, emissions, and employment. These findings leave open the possibility of relationships that this study cannot estimate precisely. They also caution against comparing the two primary analyses as direct replications: renewable share differs from physical wind capacity, and absolute changes in emissions differ from logarithmic changes.

Financing continuity offers a plausible interpretation to investigate. Resilient banks may sustain lending through difficult periods, but a higher Tier 1 ratio can also reflect a reduction in risk-weighted assets. These country-level models do not observe the lending decisions needed to distinguish those pathways. Şahin and Arzova similarly find different green-growth associations for banking stability and domestic credit [5], while Ozili and Iorember report relationships that vary with the sustainability outcome and conditioning factors [6]. Together, these findings support the study of financial resilience and the allocation of finance as related but distinct questions.

Differences from earlier environmental studies also reflect what each study measures. Hoang reports associations between financial stress and environmental outcomes that vary across countries and conditional quantiles of environmental outcomes [7]. Javed and colleagues examine whether a macro-financial risk rating conditions the relationship between financial development and green growth in developing economies [8]. The present study adds employment to the EU comparison and distinguishes market stress from bank capital and loan quality. Across these settings, the choice of financial measure and sustainability outcome is central to the interpretation.

The supporting index illustrates both the value and the limits of aggregation. Its negative annual-change association with CLIFS passes the six-test adjustment but not the broader 57-test adjustment. Employment accounts for approximately 57.3% of the weighted point estimate in the index decomposition. That share has no reported confidence interval and is neither a causal contribution nor a proportion of explained variation. Fixed reference values and explicit weights make the construction transparent, as recommended for composite indicators [20], but the index permits improvement in one component to offset deterioration in another. It also omits aspects such as job quality, distribution, and environmental conditions beyond the selected energy and emissions measures. The exploratory nonlinear tests found no adjusted evidence for the specified curvature or fixed-median threshold; this does not rule out other nonlinear relationships.

The fixed-effects design describes conditional within-country relationships. Changing national conditions may affect both finance and sustainability, and the timing results show why baseline-year stress requires particular care. Country-clustered inference accounts for dependence within countries but still relies on assumptions about dependence across them [18]. Concentrated capital variation and differences in reporting perimeters further affect precision and comparability. These features guide the next research step: combining identified financial disturbances with project, firm or lending data to examine how financing continuity connects financial resilience to investment and employment. The present evidence identifies a recurring stress–employment pattern and an environmental relationship that remains uncertain.

6. Policy Implications

Financial and sustainability monitoring benefit from keeping capital adequacy, loan quality and market stress visible as separate indicators. The recurring negative stress–employment estimates make the CLIFS measure relevant to discussions of social adjustment alongside labour-market information. The primary estimate does not reach statistical significance after the three-outcome adjustment, and the earlier-stress check changes its direction. CLIFS is therefore a contextual indicator in this study, rather than a validated employment predictor or intervention trigger. Decisions about capital requirements need evidence on their own effects; these associations do not establish such effects.

Sustainability reporting should preserve the component outcomes behind an aggregate score. In this study, employment accounts for most of the negative CLIFS point estimate in the supporting index; the index does not establish a general environmental association. Reporting employment, renewable energy share, net wind capacity, and emissions separately would show which outcome drives the aggregate result. Renewable share and wind capacity should also retain their distinct definitions, since a change in the energy mix does not necessarily indicate an equivalent change in generating capacity. The index can summarise a stated set of measures, but should not substitute for them in assessing policy progress.

For investment and labour policy, the findings help identify where more direct evidence would be most useful. Project and lending records could reveal whether financing interruptions delay transition investment and which projects are affected. Linked firm and worker data could show whether employment adjustment comes through hiring, job losses or changes in participation. Evaluating policies using those data would link financial conditions to decisions that aggregate country indicators cannot capture.

7. Conclusions

Does financial system stability foster sustainability? In these EU country panels, the answer depends on the financial dimension, the outcome, and the strength of the required inference. In the primary financial market-stress analysis and the exploratory common-sample models, higher financial market stress is associated with smaller annual changes in the employment rate. However, the primary association is not statistically significant after adjustment for the three outcomes, and measuring stress before the baseline year gives a positive but imprecise estimate. The environmental estimates and primary Tier 1 results do not establish a consistent favourable relationship with the outcomes studied. For wind capacity, the primary and bootstrap tests reach different conclusions about statistical significance. Overall, the evidence points to a recurring stress–employment pattern, while a broader environmental benefit from greater financial stability remains unestablished.

The study contributes by bringing capitalisation, asset quality and market stress into a differentiated account of environmental and social change. A supporting sustainability index reproduces the stress pattern largely through employment, thereby showing why an aggregate association should be interpreted with caution. The common-sample comparisons add evidence under outcome-specific controls, yet remain exploratory and sensitive to broader adjustment. Taken together, the results highlight a recurring stress–employment pattern, while the environmental associations and the evidence for Tier 1 capitalisation and loan quality remain less conclusive.

This distinction matters for assessing progress. A financial indicator may accompany changes in employment without showing a clear relationship with environmental outcomes. Reporting the components makes that pattern visible and keeps the interpretation of an aggregate sustainability score tied to what actually changes.

Further research can address this boundary in three focused ways. Designs with externally identified financial disturbances can distinguish stress responses from conditions already developing in the baseline year. Project, firm and lending data can connect national financial conditions to the allocation and continuity of transition finance. Broader social and environmental measures can test whether the employment-centred pattern extends to job quality, distribution and other ecological outcomes. The policy lesson from this comparison is to assess financial resilience, transition investment, and social adjustment together while preserving their differences.