Climate risks, broadly classified as physical and transition risks, play a crucial role in shaping asset price volatility and consequently, systemic risk in banking (Volz et al. 2020). Physical risks arise from climate-related events such as extreme weather, rising sea levels and changing rainfall patterns, directly impacting business operations and supply chains (Battiston et al. 2021). Transition risks stem from the economic shifts required to transition to a low-carbon economy, affecting industries reliant on fossil fuels, regulatory policies and investor preferences (Zhai et al. 2023). These risks manifest across macro-, meso- and microeconomic levels, amplifying market risk, liquidity constraints and financial instability (Bolton et al. 2020; Sato, Tasca & Isogai 2019; Wu et al. 2023), making climate change a significant driver of systemic risk in banking.

The theoretical framework linking climate change and bank systemic risk is grounded in traditional asset pricing theories including EMH (Fama 1970) and AMH (Lo 2004, 2017). Fama’s EMH (1970) suggests that asset prices fully integrate all available information, implying that climate risks should be immediately priced into financial markets. In the context of systemic risk, an efficient market would, in theory, signal any impending financial instability as a result of climate change by adjusting asset prices accordingly (Yusuf, Araoye & Afolabi 2024). However, given the uncertainty and long-term nature of climate risks, markets often struggle to fully incorporate these factors, leading to mispricing (Naseer et al. 2024). This failure can result in financial institutions, particularly banks exposed to climate-sensitive sectors, underestimating risks until a systemic crisis unfolds (Battiston et al. 2021; Bolton & Kacperczyk 2021).

The AMH extends from the EMH by proposing that markets adjust and evolve as investors respond to new risks and competitive forces (Lo 2004). Applied to climate change, this framework suggests that market participants incrementally refine their understanding of climate risks, integrating them into pricing and decision-making processes. However, cognitive biases and the unpredictable nature of climate-related disruptions may slow this adjustment, leading to mispricing and potential financial instability (Lo 2017). Complementing this is the financial contagion and risk transmission framework, which explains how climate-related shocks propagate through the banking system (Battiston et al. 2021). Forbes and Rigobon (2002) argue that what appears to be contagion in stock markets may be the result of inherent interdependence among these markets. In the context of climate risks, localised disruptions such as severe climate-related disasters and abrupt policy changes can trigger stress in one segment of the financial sector, which may then spread through these interlinked networks. Supporting this notion, Raddant and Kenett (2021) illustrates how market disruptions during the GFC underscored the vulnerabilities inherent in interconnected markets. These studies collectively suggest that even when individual financial institutions employ robust risk management practices, the interconnected nature of the financial system can amplify the adverse effects of climate-induced shocks, thereby escalating systemic risk.

These theories provide an integrated framework to understand how climate-driven asset volatility influences systemic risk. Market inefficiencies, adaptive investor behaviour and contagion dynamics amplify volatility, affecting financial stability, particularly in climate-vulnerable economies like South Africa.

Numerous studies have investigated the relationship between climate change and financial stability, with research spanning multiple regions, including the global, Eurozone, United States and Chinese contexts. For instance, Wu et al. (2023) found that temperature deviations negatively impact financial stability in China, with delayed effects and regional disparities. Similar studies in China found that climate change exacerbates financial instability, with Meng, Wang and Ding (2023) showing that carbon tax policies enhance stability over time, while An et al. (2022) found that climate risks pressure markets, threatening financial stability. Wu et al. (2023) further highlight climate deviations, particularly temperature changes, weaken financial resilience and increase market instability.

Addressing methodological gaps, Battiston et al. (2021) emphasised the unique characteristics of climate-related financial risks, such as deep uncertainty, non-linearity and endogeneity. The authors advocated for innovative approaches, including network modelling and dynamic macroeconomic frameworks, to better understand the macrofinancial implications of climate change. Their work provides a foundation for central banks and financial supervisors to integrate climate risks into their policies and risk assessments, ensuring a more comprehensive approach to financial stability.

On a global scale, Mandel et al. (2021) developed a model assessing how climate-induced shocks, such as floods, propagate through financial networks, finding that systemic risk depends on countries’ exposure to natural hazards and financial leverage. Monasterolo and De Angelis (2020) analysed financial market reactions to the Paris Agreement, showing a decrease in systematic risk for low-carbon indexes, while carbon-intensive assets remained largely unaffected. Giuzio et al. (2021) similarly examine the impact of climate risk on financial stability in the Eurozone banking sector. The study finds that climate-related financial risks amplify systemic risk, primarily through credit deterioration and market volatility, underscoring the need for climate-adjusted stress testing and regulatory interventions. Dietz et al. (2016) found that high-emission scenarios significantly reduce global financial asset values, linking climate risk to asset depreciation and financial instability. Expanding on this, Dafermos et al. (2018) show that climate change increases bank leverage, firm defaults and financial instability, suggesting that climate-targeted monetary policies could mitigate these risks.

While the relationship between climate change and financial stability has been widely studied, few studies have specifically examined the effect of climate change on systemic risk. Building on the discussion with the same geographic sphere, this section isolates systemic risk as a distinct area of focus, exploring the growing body of literature that investigates how climate change amplifies systemic risks. By detangling systemic risk from broader financial stability concerns, this discussion highlights the unique mechanisms through which climate change threatens the stability of interconnected financial networks.

Wu et al. (2024) investigate the relationship between climate risk and systemic risk in the global banking sector, analysing how different transmission channels amplify financial vulnerabilities. Using cross-country data from 120 countries between 2007 and 2020, the study finds that higher climate risk exposure significantly elevates systemic risk in banks, primarily through deteriorating credit quality rather than asset devaluation. The results also suggest that banks with stronger profitability and capital adequacy are better positioned to mitigate climate-induced financial instability, underscoring the importance of resilience-enhancing policies in the banking sector.

Wu et al. (2023) examine the impact of climate change on systemic risk in the banking sector, using quarterly panel data from 16 listed commercial banks in China. The study employs a CoVaR (Conditional Value-at-Risk) approach to measure systemic risk and uses temperature and precipitation data as proxies for climate change. The findings reveal that climate change significantly increases systemic risk spillovers, with non-state-owned banks being more vulnerable than state-owned banks. The study highlights the differential effects of climate change on systemic risk across bank ownership types, emphasising the need for tailored risk management strategies to address these vulnerabilities.

Curcio et al. (2023) focussed on the systemic risks linked to extreme climate events in the financial sector in the banking and insurance sectors of the United States. The study found a positive relationship between climate risks and systemic instability, particularly under more intense weather conditions. The use of green indexes was found to reduce systemic risks more effectively than brown indexes, suggesting an advantage for sustainable policies in risk management. The potential of climate events to drive systemic risks reflects a need for more research into how these impacts shape financial stability across interconnected sectors.

Aevoae et al. (2023) analysed the role of ESG factors in assessing the systemic risk posed by climate change in the banking industry. The study, which utilised data from 67 publicly listed banks across 47 countries between 2007 and 2020, found a negative correlation between ESG factors and bank-specific systemic risk. This suggests that prioritising environmental, social and corporate governance factors can mitigate specific banking risks, helping reduce the interconnectedness of banks and promoting broader financial stability.

In a 2020 study, Liu et al. (2020) investigated the impact of climate-induced risks on systemic banking risk in China, focussing on the relationship between rising temperatures and financial stability. Using temperature data as a proxy for climate change over the period of 2005–2018, the study found that higher temperatures were associated with increased systemic risk exposure in the banking sector. This suggests that climate-related physical risks, such as extreme heat, can amplify vulnerabilities within financial systems, particularly in regions heavily reliant on climate-sensitive industries. Building on this research, Song and Fang (2023) further explored the link between temperature shifts and systemic risk in China’s banking sector. Employing the CoVaR methodology to measure systemic risk, the study analysed data from 2010 to 2020 and found a positive association between temperature variability and systemic risk. These empirical findings reinforce the notion that climate change poses significant threats to financial stability, particularly in economies with high exposure to climate-sensitive sectors. Together, these studies highlight the growing importance of integrating climate risk assessments into financial stability frameworks, especially in regions like China where climate change impacts are increasingly pronounced.

Despite growing interest in the relationship between climate change and financial stability, several critical gaps remain in the literature. Firstly, while studies such as Aevoae et al. (2023), Wu et al. (2023), Song and Fang (2023), Liu et al. (2020) and Curcio et al. (2023) have begun to explore climate change and bank systemic risk, empirical research in this area remains limited. Moreover, a review of existing research reveals a significant gap in studies examining the impact of climate change on both financial stability and systemic risk in South Africa or the broader sub-Saharan African region. While global and multiregional studies provide valuable insights, the lack of localised research limits our understanding of how climate risks impact financial systems in regions with unique economic structures, such as South Africa, which is widely recognised as highly vulnerable to the impacts of climate change.

Secondly, climate change is viewed as a peripheral issue in financial stability literature and this limits understanding of its role in systemic risk literature. Moreover, financial stability literature relies on broad indicators like the FSI which, while useful, aggregate multiple financial risks and fail to isolate the specific impact of climate risks on systemic risk within banking systems (Diallo & Ouoba 2024; Ozili & Iorember 2024; Park & Mercado 2014). Furthermore, many studies use single-metric proxies like climate deviation and carbon emissions, which may not fully capture the impact of climate change (Song & Fang 2023; Zhang et al. 2021). Conversely, broader measures like ESG indicators (Aevoae et al. 2023) lack specificity and standardisation, limiting their effectiveness in assessing climate risk. Lastly, few studies establish transmission mechanisms through which climate risk affects systemic risk, despite evidence suggesting that climate change influences systemic risk through channels such as asset volatility, credit risk exposure and financial contagion (Battiston et al. 2017).

Given South Africa’s high exposure to climate risks, this study is both topical and relevant. Our research addresses these gaps by examining asset volatility as a mediating factor, leveraging a localised dataset for South Africa, and applying the CES model, which allows for the aggregation of bank-level systemic risk, a key variable required for this study but one that alternative approaches such as MES and SRISK do not permit (Manguzvane & Ngobese 2023). This provides a more precise understanding of how climate risk exacerbates bank systemic risk. By addressing these gaps, the study seeks to contribute to both the global discourse and the development of region-specific policies and risk management strategies tailored to South Africa’s financial system.

This study follows closely the work of Wen and Ye (2014) and Wu et al. (2023). To examine the impact of climate change on bank systemic risk specified as (Equation 1):

where systemic risk (SR) represents the dependent variable and climate change (CC) is the main predictor variable in the model. Gross domestic product growth (GDPG), interest rates (IR), domestic credit growth (DCG) and exchange rates (EX) represent macroeconomic control variables used in the model. Bank capital to total assets (BCTA), loan-to-deposit ratio (LDR), return on assets (ROA), net interest margin (NIM) and the bank concentration ratio (BCR) represent the bank-specific control variables used in the model.

This study utilises quarterly data for South Africa, spanning from Q1 2000 to Q1 2020. The primary variables of interest include SR, CC indicators, asset volatility and a set of control variables. The data are sourced from credible databases, including the IMF, S&P Capital IQ, SARB for economic indicators and financial data, and CC metrics derived from the ND-GAIN website.

The rest of the explanatory variables were selected based on their significance in explaining SR and asset volatility, data availability across BRICS and insights from both theory and empirical literature (Adenuga et al. 2021; Schleer & Semmler 2015; Song & Fang 2023; Wu et al. 2023, 2024).

This article followed all ethical standards for research without direct contact with human or animal subjects.

Table 2 displays the empirical findings of the baseline model, assessing the impact of CC on bank SR using the BMA approach based on stepwise regression. All models, as shown in Table 2, show that the mean coefficient of CC is positive and remains significant across all models, even as additional explanatory variables are introduced. This suggests that CC is a consistent and influential driver of SR within the banking industry. Each variable’s mean coefficient estimates its impact on SR, with relevance measured by Posterior Inclusion Probability (PIP). A PIP above 90% indicates high significance, while credible intervals excluding zero confirm statistical significance at the 95% level.

Table 2 shows a positive and statistically significant relationship between CC and bank SR in Model 3, with a mean coefficient of 2.91. This suggests that as climate-related risks intensify, SR increases. The findings align with Wu et al. (2023), reinforcing that CC is a key driver of SR within the banking sector. The PIP for CC is 0.96, reinforcing that it is an important factor in most models generated by BMA.

The findings indicate that macroeconomic variables such as economic growth, IR and exchange rates significantly influence SR. Economic growth exhibits a negative and statistically significant relationship with SR with a negative coefficient of (0.73), suggesting that higher economic expansion mitigates financial instability. This is consistent with Wu et al. (2024) and Liu, Sun and Tang (2021), who found that higher GDP growth stabilises financial systems by improving borrower repayment capacity and reducing default rates. Conversely, IR shows a positive and statistically significant relationship with SR, reinforcing the notion that higher borrowing costs elevate credit risk, thereby amplifying systemic instability. Our results align with the findings by Tian and Li (2024) and Wu et al. (2023, 2024), who found that rising IR exacerbate financial fragility by increasing the cost of capital and reducing firms’ ability to service debt. Exchange rate volatility also significantly affects SR, with depreciation increasing financial stress, as demonstrated by Della Corte et al. (2022), who found that currency depreciation heightens credit risk exposure in emerging markets.

Among the bank-specific factors, the NIM is negative and significant, suggesting that higher bank profitability enhances resilience to SR, aligning with Song and Fang (2023). In contrast, DCG is positive and significant, indicating that excessive credit expansion amplifies SR, consistent with Tian and Li (2024) and Wu et al. (2024). Similarly, higher bank concentration increases SR through interconnectedness and contagion effects (Wu et al. 2024). The bank capital-to-assets ratio is also positive and significant, implying that higher capitalisation does not necessarily mitigate SR but may encourage riskier lending, as observed by Wu et al. (2023). This challenges the Basel III framework, which advocates for higher capital buffers to enhance financial stability (Basel Committee on Banking Supervision [BCBS] 2011). Loan-to-deposit ratio and ROA are insignificant, indicating no measurable impact on SR. Instead, broader structural factors like credit expansion and profitability are more relevant. While some studies (Song & Fang 2023; Wu et al. 2023) link loan-to-deposit ratios to banking risks, our findings show they are not significant in South Africa’s context.

Overall, the results demonstrate that CC is a significant and economically meaningful driver of SR, with a larger impact than traditional macroeconomic factors like economic growth and IR as well as bank-specific variables. This suggests that while traditional macroeconomic and bank-specific variables contribute to SR, CC is a more substantial driver. The greater magnitude of climate risk effects can be attributed to its long-term, structural and systemic nature, which differentiates it from cyclical macroeconomic fluctuations (Volz et al. 2020). Unlike traditional factors, climate risk introduces persistent disruptions, policy-induced financial volatility and asset revaluations that have far-reaching consequences beyond short-term economic cycles (Bolton et al. 2020; Dafermos et al. (2018); Scott, Van Huizen & Jung 2017). This supports the findings by Tian and Li (2024), who highlight that climate risks are not only persistent but also magnify existing financial vulnerabilities. Consequently, policymakers and financial institutions must prioritise climate risk as a central component of SR management, as its impact increasingly outweighs traditional economic and financial determinants of banking stability.

Table 3 presents the predictive performance of the BMA models which were evaluated using the LPS, which measures the out-of-sample accuracy. Detailed results for both BMA Model 4a and 4b are provided in Table 4. Among the three models compared the BMA Model 3 from Table 2, with flexible priors, BMA Model 4a and BMA Model 4b.

BMA Model 3 with flexible priors performed best, yielding the smallest mean LPS of 1.12. This suggests that the BMA Model 3 is the most reliable model for predicting SR in future scenarios, indicating that a balance between flexibility and model parsimony is important. The model with informative priors, BMA Model 4a showed a slightly higher mean LPS of 1.21, suggesting that while informative priors helped focus on key predictors, they did not improve overall predictive accuracy. The fully specified model, BMA Model 4b, which included all predictors, had a mean LPS of 1.13, showing that including all variables does not necessarily enhance predictive accuracy. These findings highlight that models that focus on key drivers without overfitting can offer superior predictive performance when forecasting SR, especially when driven by factors like CC and economic conditions.

Asset volatility was incorporated to explore the intermediary mechanism. As noted in the methodology section, we employed two models aligned with Equation 2 and Equation 3, representing steps 2 and 3, respectively. Step 2 results are shown in BMA Model 5 while BMA Model 6 displays the third step’s mediation outcomes in Table 5.

Model 5 regresses AV as the dependent variable on CC and other control variables such as economic growth, IR and bank-specific factors. The objective here is to understand how CC and other factors impact the volatility of asset prices within the banking sector. On the other hand, Model 6 regresses SR as the dependent variable on CC, AV (which is included as a mediator) and the same control variables. This model aims to explore how both CC and AV, along with other factors, influence the overall SR in the banking system.

Table 5 shows that the coefficient of CC is significant in both models. In Model 5, CC has a positive and significant coefficient of 3.90, indicating that an increase in CC significantly heightens AV. This suggests that greater climate-related risks lead to increased volatility in the value of assets held by banks. This finding aligns with findings by Wu et al. (2024) supporting the notion that CC can create economic disruptions by affecting energy, agriculture and real estate sectors, reducing asset values. Additionally, the transition risks posed by shifting policies, regulations and market preferences because of climate concerns could further harm the valuations of firms in carbon-intensive industries (Barnett 2019; Battiston et al. 2021). Investor behaviour also plays a role as short-term fluctuations in climate risks can drive market pessimism, amplifying downside risks in the stock market, and thereby increasing AV (Lu et al. 2019; Monasterolo & De Angelis 2020).

In Model 6, the coefficient for AV is positive and statistically significant, suggesting that higher AV is associated with increased SR in the banking sector. As AV rises, SR within the industry escalates, aligning with findings by Wu et al. (2023) and Bartram and Bodnar (2009). These results reinforce the idea that fluctuations in asset values can destabilise the banking system, making it more vulnerable to broader financial risks (Bolton et al. 2020; Dafermos et al. 2018). In the same equation, CC has a positive and significant coefficient of 2.12, highlighting that CC directly contributes to SR in the banking sector. As both variables are significant, the results, therefore, confirm that AV partially mediates the relationship between CC and SR. This means that CC affects SR both directly and through AV.

In this section, we conduct a robustness test to validate the reliability of our baseline model, which examines the relationships between CC, AV and bank SR. To ensure that model-specific assumptions or particular estimation techniques do not drive the baseline findings, we utilise both the Structural Equation Model (SEM) and the Baron and Kenny (1986) approach for mediation analysis.