Featured Product

    Uncovering Climate Hazard Concentrations in Loan Portfolios: A Case Study

    November 2022

    Uncovering Climate Hazard Concentrations in Loan Portfolios: A Case Study


    New sources of concentration risk in loan portfolios stem from the exposure of counterparties and their facilities to climate hazard events. Traditionally, these exposures have not been captured by standard systemic risk factors in credit and market risk models. 

    Our novel framework quantifies the concentration risks in loan portfolios due to direct exposures of counterparties to tropical cyclones, storms, droughts, floods, and heatwaves. This approach enables financial institutions to better measure and manage these risks and thus meet regulatory, disclosure, and internal risk management needs. The framework has broad business applications that range from capital impact analysis to climate hazard resilient limit setting and climate hazard event scenario analysis.

    Leveraging Moody’s proprietary and granular data on firms’ global facility locations, we capture the localized nature of climate hazard events and provide deeper and broader insights into financial institutions’ loan portfolios.

    The analysis of climate hazard concentrations can be enhanced by incorporating exposures of firms' suppliers to hazard-prone locations. Our research shows that such indirect risk is also material for firm valuation and capital allocations.  

    Responding to current challenges

    In recent years, financial regulators, international organizations, and market participants increasingly recognize climate change as a significant risk to the global economy and financial system. Absent further climate policy changes, global warming is forecasted to exceed 2°C above pre-industrial times by 2050, the temperature increase limit enshrined in the Paris Agreement.  As the climate warms, climate hazard events, such as tropical cyclones, floods, and droughts, become more frequent and severe, constituting what is known as the acute physical risks of climate change. Recent region-wide catastrophes have given us a glimpse of these future risks. The summer 2022 drought in Europe was in some regions the worst in 500 years.  In summer 2021, flooding in Western and Central Europe was estimated to have cost between €5 billion and €6.5 billion in insured losses. 

    Despite the need to quantify and analyze physical climate risk in their loan portfolios, financial institutions have yet to sufficiently incorporate appropriate methods in their risk management models. However, the European Central Bank (ECB) expects institutions to monitor and manage credit risk in their portfolios, including credit risk concentrations stemming from climate-related and environmental risks. Institutions should also adopt granular approaches to physical climate risk using detailed geolocation data.  The Prudential Regulation Authority proposes a range of tools and metrics that should be used to monitor exposures to climate-related risk factors, which could result from changes in the concentration of firms’ investment or lending portfolios.  In addition, climate risk may aggregate over time across portfolios, which makes understanding concentrations all the more important.  Other regulators point out that, although a bank might know the location of a borrower or vendor, it may not know the locations of specific assets, which is critical for measurement of physical risks.  Existing sectoral classifications may also offer an incomplete view of risks, because they ignore commonalities in climate sensitivities across otherwise distinct sectors. The International Association of Credit Portfolio Managers (IACPM), in its 2022 study, reports that organizations increasingly recognize the importance of climate risk factors in portfolio management and are accelerating efforts to include climate concentration risks into their models. 

    In this article, we illustrate an innovative framework to capture new sources of risk concentrations fundamentally driven by exposures to climate hazards. Climate hazard risk is a geographically localized problem that has long been overlooked by standard risk management models. Our framework is unique in that it allows financial institutions to look beyond the broad-brush information on climate hazard risks in their loan portfolios. With our modeling suite, we are able to answer such nuanced questions as: “Where are the counterparties in your portfolio located?” “How climate hazard prone are these locations?” “How much additional risk do the counterparties exposed to the same hazard-prone locations introduce into your portfolio?” and “How diversified is your portfolio from the climate hazard risk perspective?”

    The challenges organizations are facing are clear and manifold. Due to their low geospatial resolution, traditional credit risk models do not properly address the granularity of climate hazard risk. Relatedly, in our empirical research, we find strong evidence that localised climate hazard risk factors add another orthogonal dimension orthogonal to the country-industry decomposition often used in existing frameworks. In other words, the standard models do not capture heterogeneous exposures to climate-related hazard events across firms. As a result, they fail to translate the impact of such events on firms’ credit quality and, perhaps most alarmingly, within the portfolio context. They also fail to quantify common exposures to future climate-related hazard events across counterparties in the portfolio.

    Our framework relies on Moody’s proprietary data on firms' locations and their exposures to climate hazard events. We also leverage the results of our unique 2020 empirical study that translates firms' physical risk exposures into credit quality changes.  In the event study, we document statistically and economically significant changes in the valuation of firms impacted by historical climate hazard events.

    Our climate hazard concentration risk framework is the first of its kind and helps fill the gap in current risk management frameworks for financial institutions, contributing to their soundness.

    A practical example

    To illustrate our solution, we use sample data from various Moody’s Analytics and Moody’s ESG databases. This allows us to capture the realizations of climate hazard events in a selection of over 400 cities worldwide, covering Africa, Asia Pacific, Europe, Latin America, North America, and Middle East. We focus on the combined but decomposable impact of tropical cyclones, storms, droughts, floods, and heatwaves.

    Our framework enables financial institutions to identify firms in their portfolios that share facility locations in either the same city or a group of closely neighbouring cities. Apart from the need for granularity to pinpoint these clusters of firm facilities, our research suggests that firms may have large facility shares in locations outside the country where they are headquartered. Indeed, 20% of firms in our sample portfolio have a larger facility share in cities located outside of their headquarter country.

    Our global sample portfolio comprises a selection of public firms and their facilities. We set a homogeneous risk profile across the portfolio to highlight the climate hazard impacts, differentiating firms solely by their facility share. Figure 1 visualizes the geographical distribution of the portfolio based on the firms’ facility locations and identifies several clusters. For instance, firms’ facilities are markedly present in East and Southeast Asia, specifically in Seoul, Manila, and Bangkok.

    Figure 1 Geographical distribution of firms’ facilities in global sample portfolio. 

    geographical distribution firms facilities map

    The geographical distribution directly informs the level of exposure to different hazard types. For instance, tropical cyclones have their highest prevalence in East and Southeast Asia, the Caribbean, and the U.S. Gulf Coast, and thus these regions tend to generate most of the tropical cyclone risk in portfolios. Importantly, climate hazard risks in neighbouring cities should not be assessed independently, because climate hazard events may have a large geographical spread. For instance, our portfolio has a significant footprint in South Korea. Although counterparties' facilities are spread across South Korea, they are likely to incur losses from climate hazard events in a dependent manner due to the small size of the country and the scale of the events themselves. This constitutes an additional source of risk.

    To compare the financial risk from climate hazards across firms, we define a new risk metric: Financial Climate Vulnerability. This 0–100 scaled metric follows a granular bottom-up approach, combining the following information on firms’ exposure to climate hazard events:

    1. Climate hazard propensity in firm’s facility locations;
    2. Sensitivity of firms’ credit quality to climate hazard exposures;
    3. Correlations of climate hazard exposures between firm’s facility locations. Firms with more geographically clustered—and hence more correlated—facility exposures are likely to be more vulnerable.

    Financial Climate Vulnerability encompasses the risk introduced by firms’ current exposure to climate hazard events. By incorporating correlations between firms, the metric can be assessed at the segment- or portfolio-level.

    Given the first component above, a natural tool for managing the climate hazard exposure in a portfolio is segmentation by primary city of hazard exposure. Table 1 shows the top five weighted cities in the portfolio, alongside Portsmouth, one of the less exposed cities. We study storm, flood, and tropical cyclone as the firms within these cities have a low exposure to droughts and heatwaves. 

    Table 1 Primary city segmentation

    Primary city segmentation

    Aggregating the portfolio in this way provides a useful tool for ensuring it does not have substantial holding amounts in heavily hazard exposed segments. In our sample portfolio, we observe, from Table 1, a large notional weighting in hazard-prone cities. The Manila segment contains one fifth of the total notional weight and stands out as the most vulnerable, mainly due to its high tropical cyclone exposure. Firms located in other cities—such as Tokyo or Portsmouth—are generally less vulnerable to climate hazards and offer attractive options for portfolio diversification.

    Financial Climate Vulnerability of the firms in the portfolio can be also aggregated at a coarser geographical level to identify the risky groups of cities. Figure 2 shows the average Financial Climate Vulnerability of firms by their primary location at the regional level and decomposes it into individual hazard types. In our portfolio, firms in East and Southeast Asia are most vulnerable, primarily due to their exposure to tropical cyclones and floods.

    Figure 2 Average Financial Climate Vulnerability by firms’ primary locations grouped by region and hazard composition. 

    Average Financial Climate Vulnerability graph

    Next, we go beyond the firm-level analysis and study how firms in the portfolio relate to each other. The granular approach of our framework allows firms with nearby facilities to share climate hazard exposures. Such common exposures make firms more likely to exhibit co-movements in their credit quality due to the impacts of the same climate hazard events. As discussed in the introductory section, these shared effects on credit quality would not be picked up in a standard credit or market risk model as they are orthogonal to traditional systemic risks, such as country-industry factors. In essence, we add a new source of portfolio concentration risk.

    Two key factors influence the climate hazard portfolio concentration risk: firms’ Financial Climate Vulnerability and facility clustering. Facility clustering is an index that measures the relative proximity of the firm's facilities to the rest of the portfolio. Figure 3 plots these two firm-level metrics against one another. Other things being equal, firms with higher vulnerability and closer proximity to facilities of other firms contribute more to the climate hazard concentration risk, an observation apparent in the top-right section of the figure.

    Among the high-risk group, we identify a cluster of the riskiest firms and mark them in red. These firms are located predominantly in the Asia-Pacific region, with about 70% of facilities clustered in Manila and Seoul, some of the most climate hazard-prone cities in the portfolio. With the high degree of facility clustering and Financial Climate Vulnerability, these firms are the top contributors to the climate hazard concentration risk in our portfolio.

    Figure 3 Financial Climate Vulnerability and facility clusters in the portfolio. Each dot represents a firm in the portfolio.

    Financial Climate Vulnerability and facility clusters

    Other business applications 

    This section discusses business use cases and describes how the climate hazard concentration risk model can be utilized in risk management frameworks. Each application is different and consequently leverages different tools developed to conduct the respective analyses – the climate hazard concentration risk model is embedded in these tools. We use the same sample of firms described in the previous section.

    Capital allocation  

    The severity and frequency of climate-related hazard events have been increasing substantially around the world, leading to multi-billion-dollar losses and business disruptions. The U.S. experienced 86 climate disasters with damages in excess of $1 billion in the past five years, with combined damages totaling approximately $450 billion. In 2022, Hurricane Ian is estimated to have cost $67 billion in damages alone. Financial institutions’ capital buffers might not be sufficient to withstand the losses caused by such events.

    In this section, we illustrate how our framework can be used to quantify Climate Hazard VaR—or capital—which is precisely the capital buffer needed to absorb losses introduced by climate hazard concentrations. In Figure 4, we incorporate the capital allocation results into our Financial Climate Vulnerability / facility clustering graphic and obtain a clear positive relationship. In particular, the Climate Hazard VaR values follow the pattern of risk level outlined in Figure 4. The high-risk segment contains individual Climate Hazard VaR values of up to 200 basis points (bps), and the firms within this group contribute 78% of the total collectively.

    On the chart, we highlight four firms all headquartered in East Asia. Despite being from the same region, the firms differ considerably in their geographical footprint. Facilities of firms 2 and 4 are mostly in Manila, capital of the Philippines, a hazard-prone country that sees, on average, 20 typhoons per year.  On the other hand, firm 3 is more spread-out, with significant operations in the U.S. and Europe, in locations with lower climate hazard risk. These factors are reflected in the capital allocations shown in Figure 4. The Climate Hazard VaR ranges from 7.6 bps for firm 3 to 170 bps for firm 2. This example further shows that climate hazard risks should be assessed based on granular firm location data rather than only the firms’ country or region.

    Figure 4 Climate Hazard VaR based on Financial Climate Vulnerability and facility clustering (top panel). Geographical distribution of East Asian firms' facilities (bottom panel). 

    Geographical distribution of East Asian firms' facilities  

    Industries map 

    Climate Hazard VaR benchmarking     

    With capital markets becoming more complex, there is an increasing interest in performance benchmarking from financial institutions, to enable a systematic assessment of the effectiveness of various investment portfolio strategies. With this in mind, as illustrated in Figure 5, we design five versions of the portfolio to help understand the ex-ante relative portfolio-level performance risk metric—the Climate Hazard VaR.

    All instances of the portfolio have an identical risk profile and holding amount, we merely alter the climate hazard exposure by adjusting the facility location distribution among counterparties. We create the two extreme cases—the “no concentration” and the “most concentrated”—by equally distributing the facilities across all locations available in the model and by assigning all facilities to Manila, respectively. In addition, we also create three portfolios in between the two extremes in which we vary the weights on Manila and keep everything else the same.

    The effects of climate hazard concentration are significant—the benchmarking exercise produces Climate Hazard VaRs ranging from a negligible 3.8 bps to a whopping 1,426 bps, a 375-fold increase. The Financial Climate Vulnerability of the portfolio rises as we increase the exposure to Manila, due to the high level of tropical cyclone vulnerability in this area, as shown in Table 1. Consequently, as the weight in the Manila area increases, so does the Climate Hazard VaR. In effect, by putting more weight on this city, we are increasing climate hazard concentration in a high climate risk location, thus subjecting the portfolio to higher probabilities of substantial hazard event impacts. If hazard events do occur, a larger portion of the portfolio will feel a detrimental effect on its credit quality.

    Figure 5 Climate Hazard VaR benchmarking for varying levels of climate hazard concentration risk.

    Climate Hazard VaR benchmarking

    Climate hazard resilient limit setting       

    Typically, financial institutions use exposure limits to set caps on investments into risky assets. To determine these measures, various factors such as target budget or sector profitability are used. In the same fashion, organizations should reflect their assets’ climate hazard concentrations when managing risk in their portfolios through a similar mechanism.

    Figure 6 shows the contributions to the portfolio Financial Climate Vulnerability for the city segments shown in Table 1, which make up a portion of the portfolio, alongside their notional weights. There is evidence that standalone notional weights do not suffice in limiting climate concentration risk, as they do not incorporate the climate risk profile of the firms within these segments. Even though the notional weight of the Manila segment is only about 20%, it still contributes approximately 60% of the risk introduced by the climate hazard exposure. To prevent segments from driving the climate risk in the portfolio, we can set a limit on the contribution to the overall portfolio vulnerability and ensure an appropriate level of diversification in terms of geographical hazard exposure.

    The choice of limit should adhere to an institution’s climate risk appetite. For illustrative purposes, we pick a limit of 15% (shown by the dashed line). This limit ensures that no segment contributes more than 15% of the total portfolio Financial Climate Vulnerability. We then translate this maximum hazard exposure-based limit into a maximum notional weighting.

    Figure 6 Segment contribution to the portfolio Financial Climate Vulnerability. 

    Segment contribution portfolio Financial Climate Vulnerability

    The contribution to portfolio Financial Climate Vulnerability is impacted by several factors. First is the notional weight of each segment within the portfolio. As can be seen in Figure 6, heavily weighted cities such as Seoul and Manila have much higher contributions to portfolio vulnerability than other cities. Second, cities highly exposed to climate hazard events are more vulnerable and thus contribute more to portfolio vulnerability than cities that are less exposed. Figure 6 shows that despite a higher notional weighting in Seoul, Manila has a much larger contribution to portfolio vulnerability. This is because tropical cyclone vulnerability is considerably high in Manila, whereas in Seoul there is only a medium vulnerability to flooding, as can be seen in Table 1. Third, correlations between cities also impact the contribution levels. Correlated, vulnerable cities will increase portfolio vulnerability more than uncorrelated, vulnerable cities, as hazard events are likely to impact the correlated cities simultaneously. A portfolio of less diversified climate hazard exposures will be riskier, and cities that are both highly correlated and vulnerable will drive the portfolio vulnerability.

    To continue with our example, a high-weighted, highly vulnerable segment in a concentrated area of the globe will have a tighter risk limit. Figure 7 plots the Financial Climate Vulnerability and the facility clustering index of each primary city in the portfolio, along with the corresponding notional limit, which ensures the segments’ contribution to the portfolio Financial Climate Vulnerability does not exceed 15%. Limits are more restrictive in the top right of the chart where segments have higher vulnerabilities and there are larger correlations with the rest of the portfolio. The tightest limit in the portfolio is for the Manila segment, where only 7.2% of portfolio notional may be allocated without breaching the 15% vulnerability contribution target. The bottom left of the chart includes cities such as Portsmouth where vulnerabilities and concentrations are much lower. These segments provide opportunities for further climate diversification—notional limits increase up to 48% for the same 15% vulnerability target.

    Figure 7 Financial Climate Vulnerability-based risk limits by primary city.

    Financial Climate Vulnerability-based risk limits

    Climate hazard-related supply chain risk  

    Assessing climate hazard risk based on the locations of firms’ facilities may be incomplete as they also can be exposed to material risk indirectly through their suppliers. For example, the 2011 Thailand floods caused over $45 billion in economic damages. With over 30% of Thailand’s annual exports being electronics, the flood caused widespread disruption to global supply chains, from hard drives to semiconductors.  Our research shows that such indirect exposures also can have a significant effect on firms’ credit quality. For instance, in our empirical research, North American firms with at least one supplier impacted by hurricanes exhibit, on average, a 1.3% deterioration in value over the first four months after the events.

    We can extend the climate hazard concentration risk analysis to incorporate the supply-chain risk based on the data from Cortera, a Moody’s affiliate. Figure 8 focuses on two firms to illustrate insights from incorporating the supply chain data. Based on the locations of their facilities, the two firms share little to no common climate hazard exposures. However, some of their suppliers are in the same areas of Florida and Texas with high hurricane risk, and the two firms may exhibit commonalities in their losses due to potential supply chain disruptions. In the portfolio context, these common exposures to suppliers in hazard-prone locations are a novel source of concentration risk. In some portfolios, the indirect, supply-chain-related impact on the Climate Hazard VaR can be as much as twice as high as the direct one, causing the overall capital number to triple.

    Figure 8 Locations of facilities and suppliers of two North American firms. 

    Locations of facilities north american firms

    Risk under a climate hazard event scenario

    Determining the level of expected losses in a portfolio under a specific climate hazard event is complex but now possible. Climate hazard scenario analysis is one of the new ways to assess the resilience of institutions to losses associated with extreme weather. It has become crucial for organizations to understand both the acute risks posed by hazard events and the long-term chronic risks generated by climate change. Severe shocks from climate hazard events are more likely to have considerable impacts on a loan portfolio in the near-term.  For instance, the ECB 2022 climate risk stress test assumed two scenarios that modelled both the impact of a severe drought and heatwave in Europe, as well as severe floods in Europe over a one-year time horizon.  Our framework also lends itself to addressing similar questions.

    To illustrate the analytical event-scenario capability, we analyze a hypothetical event similar to the category 5 Typhoon Chanthu, one of the most severe tropical cyclones of 2021. The top subplot of Figure 9 illustrates its trajectory: The event impacted the Philippines, Taiwan, China, and Japan, causing damage to firms with operations along its path. The bottom subplot shows the projected increases in firm-level expected losses, primarily determined by firms’ facility presence in the disaster area. Although the total expected loss of this portfolio concentrated in East Asia increases by about 14%, for the most exposed firms the increase can be as large as 150%.

    Figure 9 Path of typhoon Chanthu (top) and company-level expected loss effects of typhoon impact (bottom). 

    Path of typhoon Chanthu  

    facilities exposed typhoon Chanthu


    The identification of climate hazard concentration risks plays a critical new role in credit risk management for financial institutions globally. An increasing number of financial regulatory bodies and other industry standard-setters are recognizing the financial and prudential benefits of measuring and monitoring risk concentrations that stem from climate-related risks.

    Our novel framework, which combines spatial correlations of climate hazards with their impact on corporate valuations that we estimate, enables organisations to consider the new sources of concentration risks in their decision-making, including concentration management. In its core, the framework equips the user with the necessary tools to understand the physical presence of the counterparties’ and their suppliers’ facilities, how hazard-prone these locations are, the degree of hazard correlation, and how the combined effect might impact the exposures’ credit qualities and their co-movements.

    Among its further use cases, the model helps assess reductions in investment risk through portfolio construction and diversification—by calculating capital allocations and benchmarking, setting climate hazard resilient limits, running scenario analyses using both realized as well as hypothetical weather events, and assessing the connectedness embedded in supply chains.

    Print Download