This chapter outlines the advantages, challenges and policy directions related to catastrophe (CAT) bonds in Dynamic Asia and the Pacific. CAT bonds can provide a fast solution to natural disaster exposure as they do not require a pre-existing insurance market, and parametric triggers allow for a fast pay-out mechanism. In addition, CAT bonds minimise counterparty risk through full collateralisation with high-quality securities and allow for efficient price discovery through a decentralised market. Challenges that hinder the development of CAT bond markets include infrastructure prerequisites, funding constraints, and regulatory limitations. Additionally, parametric triggers lead to basis risk. Policy recommendations emphasise a holistic approach to strategic risk transfer planning, knowledge-building, infrastructure investment, data quality enhancement, and broadening investor bases.
Fostering Catastrophe Bond Markets in Asia and the Pacific
3. Benefits and challenges of developing catastrophe bond markets in Asia and the Pacific
Abstract
Introduction
Natural catastrophes claim human lives and cause substantial economic losses across the globe. The associated costs have grown continuously over the last four decades, from an annual average of USD 30 billion in the 1980s to USD 207 billion between 2010 and 2019 (Swiss Re, 2022[1]).1 Insurance often covers just a fraction of these costs.
The financial burden resulting from underinsurance, the often referred to as the disaster protection gap, is either shouldered by the public sector or remains with households and companies. The countries of Dynamic Asia and the Pacific usually exhibit larger protection gaps than advanced economies and have less financial flexibility at the government level (see Chapter 2). They thus find it much harder to deliver comprehensive disaster relief and recovery packages as well as a strong fiscal response for reconstruction in the wake of major catastrophe events.
The countries of Dynamic Asia and the Pacific have various options for mitigating this problem. The most straightforward is to promote the maturation of national insurance markets for the coverage of disaster risk. Yet even with rapid advances in technology, it will take time to build up the required organisations and capital bases. Another option is a national disaster fund that underwrites the risk from households and companies. A key challenge in this case is on the distribution side. Alternatively, to secure risk financing directly from capital markets, governments can use insurance-linked securities (ILS). The most popular and tested instrument in this regard is the sovereign-sponsored catastrophe (CAT) bond.
This chapter explores the possibility of disaster-risk transfer through catastrophe bonds in Dynamic Asia and the Pacific. It finds that CAT bonds are a promising means of improving resilience. This is mainly because these bonds can be issued in the absence of a mature insurance market, in most cases they allow for fast settlement. Nevertheless, several countries in Dynamic Asia and the Pacific will need to establish the fundamental preconditions for catastrophe bond issuance, including reliable data providers and measurement stations, suitable risk models and a transparent and robust process for allocating proceeds in case of a trigger event.
This chapter firstly considers the main benefits of CAT bonds for the economies of Dynamic Asia and the Pacific. It then discusses the major challenges that must be overcome to reap these benefits. Finally, the chapter offers policy recommendations and attempts an outlook for disaster-risk financing strategies via catastrophe bonds.
Benefits of catastrophe bonds for Dynamic Asia and the Pacific
Catastrophe bonds, financial instruments that utilise a process called securitisation to wrap natural disaster risk into a tradable format (see Chapter 2), can, in most cases, provide a fast means of absorbing the impact of natural catastrophes in the short run since bond protection can be put in place immediately. Within a risk financing strategy, CAT bonds are one of the options for ensuring adequate funding for disaster responses, and they are ideal for transferring low-frequency, high-severity risk. Moreover, catastrophe bonds can be issued anytime and have a typical term of three years, therefore offering flexibility and price stability. CAT bonds are designed to immunise the sponsor against counterparty default risk through full collateralisation with high-quality securities. Furthermore, the use of parametric triggers potentially allows for a quick source of funding, while price signals from the CAT bond market and modern pricing models allow for informed decision making. Investors need to diversify their ILS portfolios, thus they tend to accept a spread discount by sharing risks at a multi-country level.
The discussion that follows will consider various benefits of the use of catastrophe bonds:
diversification of coverage
flexible multi-year coverage
full collateralisation
transparency and, in most cases, fast settlement
efficient price discovery
multi-country risk sharing
CAT bonds are an important instrument for the diversification of coverage
The countries of Dynamic Asia and the Pacific have various options for financing disaster losses that accrue to the assets of households, companies and the public sector. The extent to which governments can leverage these tools depends on rigidities in the budget as well as competing economic, social and political objectives. Considering these pressures, this section reviews the public financial management practices that governments can use to respond to disaster shocks.
Figure 3.1 shows various financing tools that can be used to cover costs from disasters. Government’s optimal post-disaster response is determined by the type of hazard and the magnitude of disruptions to economic activity, among other things (OECD, 2022[2]). Budget reallocations are suitable for absorbing losses caused by disasters that will not overburden a country’s fiscal capabilities. When disasters cause emergencies, funds within a budget can be redistributed across programmes, budget lines and ministries thereby allowing a government to restructure expenditures within the range of its approved budget, limiting the fiscal shock. However, depending on the size of the hazards to be covered, budget reallocations can be costly as they put other objectives at risk. Shifting resources from one spending priority to another can cause disruptions in the provision of certain public goods and services.
Figure 3.1. Funding approaches to cover contingent liabilities from disasters
Source: (OECD, 2022[2]), Building Financial Resilience to Climate Impacts: A Framework for Governments to Manage the Risks of Losses and Damages.
On the other hand, sovereign CAT bonds and reinsurance coverage secured through a sovereign disaster risk fund pertain to the higher loss layers that will be reached when a rare large-scale disaster strikes. These risk tools ensure that post-disaster funding through financial support from donors, tax increases or the issuance of government bonds comes into play for the residual losses. In many cases, tax hikes and debt financing are not a country’s desirable options. Tax increases take time to implement and can have a negative impact on the economy as they reduce disposable income and consumer spending, while debt financing may affect a country’s longer-term fiscal flexibility. Although countries can hope for the financial support of donors, it takes time to mobilise international aid, and such aid typically covers only a small fraction of the overall disaster losses (Swiss Re, 2018[3]).
In light of these considerations, CAT bonds are an important instrument for the diversification of coverage in a disaster risk financing framework. Such a framework can be developed based on both theoretical guidelines and best-practice experiences. For example, Clarke et al. (2017[4]) propose a framework that supports sovereigns in evaluating ex ante risk financing instruments and choosing an appropriate combination, taking economic and political constraints into consideration. The size of a CAT bond relative to the weight of the other instruments in the framework should be determined, among other factors, based on the current price of coverage in the CAT bond market relative to the cost of alternative measures, such as traditional reinsurance coverage.
Finally, planning the funding for a disaster budget should take place early. Financing of the spreads (ROLs) for CAT bonds can be challenging for emerging market sovereigns, as they may have limited tax income, restricted access to capital markets and high borrowing costs. Hence, it is of utmost importance to plan early and integrate the costs for the whole disaster risk management programme into fiscal planning on an ongoing basis. Good fiscal planning requires the effective management of public resources. Governments must balance the funds required to pay the premiums during the term of the CAT bond against other public spending needs. However, they should also consider the present value of the expected future cost of relief, recovery and reconstruction, which would need to be funded if no disaster risk coverage is put in place.
To alleviate the strain of the spread on the country’s budget, governments should aim to exploit the non-peak territory and multi-country discounts in the ROL, discussed above. To this end, they could engage in roadshows to educate investors about the diversifying properties of non-peak perils for ILS portfolios that are heavily US-centred. They could also initiate discussions with neighbouring countries that are exposed to the same types of natural disasters. If the opportunity for a supranational risk pool arises, all member countries would be able to benefit considerably from the sharing of the issuance costs and the multi-country spread discount.
Countries have different fiscal capacity, to mobilise their budget, use debt financing and access risk financing and risk transfer markets. Different countries will therefore choose different financing solutions based on the conditions they face.
CAT bonds offer flexible multi-year coverage
CAT bonds can be issued at any time and typically exhibit a term of three to four years. Accordingly, they guarantee emerging countries multi-year coverage for natural catastrophe risk. This is not the case with traditional reinsurance contracts, e.g. those accessed through a sovereign disaster fund, because global reinsurance markets are subject to annual renewal at fixed dates2 (Gallagher Re, 2022[5]). One-year coverage requires annual renegotiation. This is time consuming and resource intensive for both cedents and risk carriers. Multi-year contracts can reduce this administrative burden. Moreover, multi-year coverage affords cedents better strategic risk management planning. Above all, however, one-year coverage exposes cedents to significant price risk because the rate on line (ROL), or the ratio of premium paid to loss recoverable, is known to fluctuate substantially over time, depending on the amount of risk capital available to the industry, among other factors (Cummins and Trainar, 2009[6]). This phenomenon, known as the reinsurance underwriting cycle, is illustrated by the Guy Carpenter Regional Property Rate-on-Line Index (Figure 3.2).
Figure 3.2. Reinsurance underwriting cycles
Note: The Guy Carpenter Regional Property Rate-on-Line Index, a proprietary index of regional property catastrophe reinsurance ROL levels, has been published since 1990 by Guy Carpenter, a global reinsurance broker. The prices are based on the firm’s brokered excess of loss reinsurance contracts.
Source: Guy Carpenter, https://www.artemis.bm/regional-property-cat-rate-on-line-index/.
Locking in a multi-year ROL through CAT bonds insulates sovereigns against these large price swings and removes uncertainty from their fiscal planning process. This is particularly valuable after the occurrence of large natural disasters, which tend to drain a substantial amount of capital from the reinsurance industry and subsequently cause a major surge in prices. Figure 3.2 shows the impact on the ROL of two of the costliest natural disasters in history, Hurricane Katrina in the United States (2005) and the Tohoku earthquake in Asia (2011). The consequences of Katrina were severe for cedents, who were abruptly forced to pay much higher premiums to maintain their reinsurance coverage. As insurers passed on costs to their customers, this led to increased rates for policy holders.
Many cedents therefore integrate CAT bonds into their risk management strategy specifically because of the multi-year coverage they offer. In 2022, the French reinsurance company SCOR successfully sponsored Atlas Capital Reinsurance 2022 DAC, a CAT bond that provided it with multi-year coverage of USD 240 million against named storms in the United States, earthquakes in the United States and Canada, and windstorms in Europe. According to Jean-Paul Conoscente, SCOR’s chief executive officer, the decision to use a CAT bond was driven by the desire for multi-year coverage (SCOR, 2022[7]).
CAT bonds help to minimise credit risk via full collateralisation
Catastrophe bonds have been designed to minimise credit risk. This is achieved through full collateralisation (Lakdawalla and Zanjani, 2011[8]). The typical CAT bond collateral consists of United States treasury bills and is therefore very safe. Accordingly, both investors and sponsors can avoid the consequences of a counterparty default. For traditional reinsurance, in contrast, some credit risk remains despite the high financial strength ratings of reinsurance companies.3 The reinsurer selling coverage to a sovereign disaster risk fund may fail to pay out following a major natural catastrophe.
If an insurance product fails to pay out following a major natural catastrophe, the effects can be devastating. In the case of an emerging country seeking coverage for the purpose of funding disaster relief and recovery, and possibly reconstruction as well, such a double default scenario would have dramatic consequences for the well-being of citizens and the financial resilience of firms. Empirical evidence shows how important this “remaining” credit risk is in reinsurance markets. For example, using data from the US property-liability insurance industry from 2002‑09, Park Xie, and Rui (2018[9]) examined how sensitive reinsurance demand is to credit risk. They found that reinsurance ceded to a counterparty reinsurer decreases if the reinsurer suffers a rating downgrade. Their results hold at a 1% significance level and show large negative effects of a downgrade, even if the previous reinsurance rating is good. Similarly, Park and Xie (2014[10]) found that both financial ratings and stock prices of ceding insurers react negatively to downgrades by counterparty reinsurers. Hence, full collateralisation is a key benefit of CAT bonds for the cedent, as it notably reduces the uncertainty pertaining to the availability of funds in the case of a trigger event.
CAT bonds have high transparency and permit fast settlement, in most cases
In contrast to other instruments, CAT bonds exhibit high transparency and can deliver rapid payout in the case of a trigger event. This is particularly true if they rely on the parametric trigger mechanism. When a natural disaster has occurred and the respective measurements of the physical parameters have been taken, the transaction can be settled in a matter of days. For instance, only five weeks after a magnitude 8.1 earthquake hit the Chiapas region of Mexico in September 2017, a FONDEN-sponsored CAT bond paid out USD 150 million (Artemis, 2017[11]). In contrast, under an indemnity-based reinsurance contract, all incurred losses must first be verified through the claims management of the reinsurance company. Similarly, raising capital after a disaster by issuing government debt or raising taxes takes a considerable amount of time. A rapid payout can help to avoid additional expenses for cedents and investors.4
As delayed payments quickly lead to distrust in the viability of a risk management strategy, they could lead institutions and the general population of emerging countries to withdraw support for the sovereign risk transfer programme with CAT bonds. This would imply a step backwards in terms of the goal of reducing global protection gaps. Payout delays can have a major impact on impoverishment and economic growth, especially for emerging economies (World Bank, 2017[12]). Since their asset base is naturally low, the bridging of payment failures is often disproportionately expensive or impossible. This effect is particularly pronounced for vulnerable populations, such as women and children, and when the insured catastrophe also affects people’s labour income.
Parametric triggers, in addition to allowing swift payout, also reduce the risk of legal disputes with investors because the parameter values are usually measured and published by an independent third party. This implies that the payout is based on a highly transparent measure that can be manipulated neither by the seller nor the buyer of protection. It should be mentioned, however, that even a parametric CAT bond does not completely exclude the possibility of disputes over the payout. Specifically, in the case of storms that impacted Mexico in 2014 (Hurricane Odile) and 2015 (Hurricane Patricia), storm chasers delivered atmospheric pressure readings that differed from the figures reported by official sources such as the US National Hurricane Center. This led to a delay of three months in the payout of Mexican CAT bonds (Blackman, Maidenberg and O’Regan, 2018[13]).
CAT bonds allow buyers efficient price discovery
Prices for CAT bonds are less opaque than those for traditional reinsurance. The existence of an over-the-counter secondary market ensures that indicative price sheets by broker dealers are updated on a regular basis. This provides prospective buyers with reliable signals on the current market pricing of natural disaster risk (Beer and Braun, 2022[14]). In contrast, reinsurance markets exhibit an oligopolistic structure. They update their pricing once a year for renewals. The up-to-date pricing information produced by the CAT bond market can be harnessed to calibrate accurate econometric and financial pricing models that have been developed in recent years (see Annex 3.A.). Moreover, machine-learning approaches for the pricing of risk have become available (Götze and Gürtler, 2020[15]; Makariou, Barrieu and Chen, 2021[16]). The combination of adequate models and frequent market data ensures reliable price tags for risk transfer that enable informed decision making.
However, prospective cedents should be aware that the efficiency of price discovery depends on secondary market liquidity, which varies over time. It should also be noted that CAT bonds are less liquid than corporate bonds (Lane and Beckwick, 2016[17]). Herrmann and Hibbeln (2022[18]) find that a seasonality‐implied increase of default risk leads to a substantial reduction of CAT bond trading, even in periods with much new information arriving in the market. The fact that CAT bonds referencing seasonal perils, such as cyclones, are less frequently traded during the risk season implies a higher likelihood of stale prices. This is different from the trading patterns of corporate bonds, where the information effect dominates (Herrmann and Hibbeln, 2022[18]). Issuance activity in the primary market for CAT bonds exhibits the same pattern: it is much lower during the hurricane season from June to November (Braun, 2015[19]). It is therefore advisable to plan the placement of coverage strategically and to avoid periods of low liquidity and inefficient price building.
Multi-country cost sharing offers advantages and a spread discount
A prominent example is the Caribbean Catastrophe Risk Insurance Facility (CCRIF), which issued a multi-country CAT bond in 2014, that provided insurance coverage for a group of sovereigns in the Caribbean and Central America against hurricanes, earthquakes and extreme rainfall events. Another example is the parametric Pacific Alliance CAT bond for earthquakes in Chile, Colombia, Mexico and Peru.5
Multi-country structures such as the CCRIF or the Pacific Alliance CAT bonds, offer substantial cost advantages (see Chapter 5 for a detailed discussion). Member countries of the pool can share structuring, legal and issuance expenses, which together are a major quantity.6 Examples are the costs for the offering circular, the catastrophe risk modelling and the placement of the bonds. If the multi-country CAT bond covers adjacent countries, it may be possible to use the same risk model for example, especially when the countries’ geological, hydrological and meteorological characteristics are similar. This is particularly useful for small developing nations with geographic similarities. A joint risk management programme with neighbouring nations affected by the same disaster risk can significantly ease the strain on fiscal budgets.
What hinders CAT bond development? Constraints and challenges
This section discusses challenges fostering catastrophe bonds in Dynamic Asia and the Pacific. When parametric triggers are used, it may lead to basis risk scenarios where a country has been struck by a disaster, but its sovereign CAT bond does not pay out. The use of parametric triggers also requires advanced and reliable infrastructure. An example is mesonets, or networks of automated weather stations, that can withstand cyclone wind speeds. Data providers that are independent from the government and adhere to the highest standards of data processing, storage and submission are also needed, together with suitable catastrophe risk models to fill the historical data void. Funding of the CAT bond spreads often poses a challenge as well. When emerging country budgets are too tight to afford the CAT bond spread, sponsors may seek support through development aid. Moreover, the sovereign sponsor needs to design efficient and fair distribution schemes and ensure that social vulnerabilities are considered. The lack of a track record in CAT bond issuance can hamper investor interest. Investors may be concerned about the illiquidity of CAT bonds from Dynamic Asia and the Pacific, and little trading activity may create pricing uncertainty. Lastly, regulatory issues are another key factor limiting CAT bond market development. In many countries in Dynamic Asia and the Pacific, CAT bonds are a relatively new financial product, and legal and regulatory frameworks remain underdeveloped.
The discussion that follows will consider challenges to the use of catastrophe bonds, including:
basis risk
measurement infrastructure
data quality
rapid and target-oriented distribution of the proceeds
liquidity and valuation concerns
inconsistency in regulatory treatment.
Parametric triggers may lead to basis risk scenarios
Policy makers in Dynamic Asia and the Pacific may be reluctant to sponsor CAT bonds, which typically involve relatively high up-front costs and a possible payoff at some point in the future. The one-time costs of issuing CAT bonds tend to be higher than those of other types of debt securities (Michel-Kerjan et al., 2011[20]). CAT bonds are typically structured using offshore special purpose vehicles (SPVs) or special purpose insurers (SPIs).
As discussed in Chapter 2, most CAT bonds issued by insurance and reinsurance companies today exhibit indemnity triggers. Public sector CAT bonds, in contrast, often rely on the parametric trigger. The reason is rather straightforward. Unless a national risk pool acts as the CAT bond sponsor,7 there is no portfolio of insurance policies that could be referenced by an indemnity trigger. The parametric trigger has further advantages in addition to its employability in the absence of an insurance portfolio, particularly transparency and rapid settlement.
The main drawback of parametric triggers is basis risk for the cedent.8 Basis risk describes the situation in which a sovereign issuer may face substantial fiscal strain in the aftermath of a catastrophe, but the CAT bond does not pay out because the parameter value measured at the relevant geographic location did not exceed (or fall below) the trigger threshold. This happened, for example, in the case of MultiCat Mexico 2012-1, a sovereign CAT bond issued by FONDEN. Hurricane Odile hit Baja California in September 2014 and caused more than USD 1 billion in economic losses. Despite this substantial damage, the central pressure of the storm simply did not fall below the predefined threshold of 932 millibars (mb) (Artemis, 2014[21]). A similar situation occurred when Hurricane Sandy struck the Caribbean in 2012. Although three CCRIF member countries were impacted severely (Jamaica, Haiti, and the Bahamas), the event did not qualify for a payout from the programme. In two of the three cases, model-based loss estimates were below the trigger threshold, and in the third, the country was located outside of the modelled wind field (Artemis, 2012[22]).
Basis risk has been shown to negatively affect the demand for coverage (Mobarak and Rosenzweig, 2013[23]). It is hence crucial that emerging market sovereigns understand the consequences of basis risk and take measures to minimise it where possible. While the basis risk inherent in parametric triggers can never be fully eliminated, it can be mitigated through parametric indices and proper catastrophe risk modelling. These options will be discussed in greater detail in the context of the policy recommendations that close the chapter.
Parametric triggers require reliable measurement infrastructure
Parametric triggers require reliable measurement infrastructure in the geographic territory covered. In the earlier days of the CAT bond market, parametric triggers for cyclone risk were not possible even in the United States due to the lack of a reliable network of hurricane-hardened weather stations. Standard anemometers, which measure wind speed and direction, needed to be deactivated or failed because of extreme storm gusts. This led to a survival bias, because only those stations that did not experience the most severe wind conditions were able to provide readings (Brookes, 2009[24]). A network of catastrophe-resistant meteorological or geological measurement stations (ideally with redundancies) is thus crucial for the accurate recording of wind speeds or earthquake strengths for parametric CAT bonds.
The measurement network should also guarantee sufficient density of stations (UN ESCAP, 2015[25]). A report by India’s Agricultural Finance Corporation showed that 77% of farmers who used parametric crop insurance were dissatisfied with the location of weather stations (AFC, 2011[26]). The stations had been installed too far from their farms and thus did not adequately reflect their exposure.9 The World Food Programme, in a joint study with the International Fund for Agricultural Development, estimated in 2010 that for accurate insurance coverage in India, the weather station network would need to be expanded by additional 10 000 to 15 000 units, implying an investment of USD 5-6 million in installation costs and an additional 25% per year in maintenance costs (World Food Programme, 2010[27]). Emerging countries that lack measurement infrastructure may therefore need a significant public investment before parametric CAT bonds can be utilised for sovereign risk transfer.
For other natural perils, such as floods or drought, satellite or drone data could be a cheaper alternative (Matheswaran et al., 2018[28]; Whitehurst et al., 2022[29]). Satellite data is cost efficient and less prone to manipulation than data from conventional weather stations. Yet there are limitations, such as reduced performance in cloudy periods and over mountainous terrain (UN ESCAP, 2015[25]).
Data quality and reliable data providers are also needed
Apart from the measurement infrastructure, sovereign CAT bonds are not feasible without trustworthy data providers. First, the private or public organisation operating the networks of measurement stations in the emerging country, e.g. the national weather service or geological science agency, needs to be independent from the government and adhere to the highest standards of data processing, storage and submission. This will ensure accurate and reliable readings for each catastrophic event that are acceptable as a basis for million-dollar CAT bond transactions. Without such independence, the government could exert pressure on the meteorological or geological service to manipulate readings from measurement stations and trigger a payout from their sovereign CAT bonds. Moreover, the budget of the service may be subject to political whims, leading to underfunding that could threaten its ability to perform its duties effectively.
Second, even if the physical parameter values themselves are measured correctly, financial market investors will not be prepared to invest their capital if the calculation agent for the CAT bond is not trustworthy. The calculation agent is the entity that generates the index or modelled loss estimate for the CAT bond based on the recorded raw data. It thus plays a crucial role in determining whether a payout is due. To avoid a major conflict of interest, the calculation agent should be an independent third party. Otherwise, it may have a vested interest in manipulating the payout calculation in favour of one party over another. For example, if the calculation agent is affiliated with the government that sponsored the bond, it may be incentivised (or pressured) to enable an undue payout and to maximise the proceeds, to the detriment of investors. This problem is similar to the classical moral hazard of the sponsor under indemnity triggers, which has been discussed at length in the academic CAT bond literature. Moral hazard is thought to arise in the context of indemnity-trigger CAT bonds. Since the payout under indemnity triggers is tied to the own losses of the sponsor, the latter has strong economic incentives to relax underwriting and claims handling standards after the coverage has been put in place. Empirical evidence on moral hazard is mixed. Chatoro et al. (2023[30]), and Braun (2015[19]) do not find moral hazard to be priced in the primary market, while Dieckmann (2010[31]), Papachristou (2011[32]), and Götze and Gürtler (2020[15]) claim to find empirical indications for ex ante moral hazard of CAT bond sponsors. Investors who doubt the independence of the calculation agent will not trust the accuracy and fairness of the payout calculation and will turn away from the CAT bond offering. The fear of moral hazard can be a particular problem for emerging market CAT bond issuances.
In addition, missing track records will be a challenge. The number of CAT bonds issued to date for sovereign risk transfer in emerging markets remains relatively small. This poses challenges to investor acceptance. While early adopters from the ILS industry have already participated in these transactions, attracting capital on a larger scale may require a longer track record and transaction history. Track records are important for investors because they provide valuable information about the past performance of an asset and the trustworthiness of the sponsor. This is particularly important in the context of exotic securities, such as CAT bonds. This issue may be relevant to pricing as Spry (2012[33]) notes that investors in the CAT bond market closely scrutinise the sponsor of a transaction.
Rapid and target-oriented distribution of the payout must be assured
Once a sovereign sponsor receives a CAT bond payout to fund its post-disaster needs, it requires sufficient personnel and processes to ensure an efficient and targeted distribution of the proceeds. Sovereign CAT bonds can be used to finance immediate relief as well as longer-term recovery and reconstruction. The first decision to be made concerns the allocation of the CAT bond funds across these three domains. This decision must consider further payouts from other components of the country’s integrated disaster risk management strategy.
The capital dedicated to relief and recovery then needs to be deployed rapidly. Relief requires temporary shelters, food, medical support, etc., while recovery entails the restoration of critical infrastructure such as electricity, telecommunications and water supply. Time is of the essence for the effectiveness of these measures. Without a high degree of organisational preparedness, it may not be possible to achieve the maximum impact. In contrast, a long-term strategy is required for reconstruction. Countries will first need to prioritise the private and public sector assets needed to rebuild. The available capital will usually not suffice to fund a full reconstruction of all damaged or destroyed assets. Hence, a distribution scheme is needed to select the households or firms most in need of government support.
Unfortunately, empirical evidence points to the fact that disaster assistance is often inequitable, both in emerging and developed countries. The reason is that the distribution of public funds for relief and reconstruction commonly focuses on damages, but not vulnerability. This is a major issue, because the most vulnerable populations are least able to cope with the consequences of natural disasters (O’Keefe, Westgate and Wisner, 1976[34]). For example, Emrich, Aksha and Zhou (2022[35]) analysed the proceeds paid to US homeowners by the Individuals and Households Program (IHP) of the Federal Emergency Management Agency (FEMA) between 2010 and 2018. They found that allocations are driven by damages while ignoring social vulnerabilities. Their study also reveals significant imbalances in terms of the ethnic and racial composition of the receiving counties. Similarly, Drakes et al. (2021[36]) investigated the relationship between short-term disaster relief and social vulnerability, based on US data. Their results showed that geographic areas with a high social vulnerability were not sufficiently served by FEMA’s IHP programme. Factoring social vulnerability into the distribution scheme for post-disaster assistance would help to combat major societal inequities.
Investors may have standardisation and liquidity concerns
Standardisation of financial instruments is a desirable feature from an investor’s perspective as it improves market liquidity and helps investors to manage their portfolios in a more efficient manner. Investors are more likely to invest in CAT bonds if they view them as relatively liquid and standardised assets that are associated with low administration costs and that adhere to transparent pricing rules (Yago and Reiter, 2008[37]; Braun, Müller and Schmeiser, 2013[38]). Although the basic structure of CAT bond follows a standardised approach, most deals are tailor-made transactions. In issuance terms, the CAT bond market is currently almost evenly split between indemnity and non-indemnity structures (Artemis, 2022[39]). This lack of standardisation partly explains the current fragmentation of the CAT bond market and may be a challenge to further growth.
The role of standardisation in CAT bond market development has been discussed in the specialised literature. For example, to explain institutional investors’ reluctance to invest in CAT bonds, Bantwal and Kunreuther (2000[40]) employ behavioural economics aspects, such as ambiguity aversion, myopic loss aversion and the fixed costs of education. The authors suggest that sponsors should aim for a larger degree of security standardisation in order to decrease demand and promote market growth. In a similar vein, (Cummins, 2005[41]) describes CAT bonds as a valuable means of portfolio diversification and emphasises that more standardised and transparent transactions, and the development of a public secondary market, would help realise the full potential of this asset class.
Illiquidity can be a major concern of CAT bond investors as it reduces the ability to trade out of an unwanted position. Illiquid securities are characterised by low trading volumes and wide bid-ask spreads. To find a buyer, investors may have to wait longer and accept a lower price. The prices of illiquid securities tend to be more volatile and may be stale. This poses difficulties in valuation and can lead to large losses if investors need to sell quickly.
Less liquid CAT bonds thus carry a higher spread. Using data from the Trade Reporting and Compliance Engine (TRACE), Herrmann and Hibbeln (2022[18]) disentangle the default and liquidity premium of CAT bonds. They find that the liquidity component included in secondary market CAT bond yields amounts to 98 basis points, or 21% of the total risk compensation, and they also document that this effect is more pronounced for high-risk CAT bonds. Although there is no empirical evidence yet for the link between the structural characteristics of CAT bonds and the illiquidity premium, it may be suspected that the illiquidity premium is even larger for countries for which no sophisticated catastrophe risk models exist. Illiquidity may not only lead to higher risk spreads but may also affect the ability to value CAT bonds in general.
Inconsistency in regulatory treatment can limit CAT bond market development
Regulatory issues are another key factor limiting CAT bond market development. In many developing countries, CAT bonds are a relatively new financial product for transferring disaster risks, and legal and regulatory frameworks consequently remain underdeveloped. Development of appropriate frameworks may allow effective risk transfer and ensure the rights of investors, increasing their confidence and their demand for CAT bonds.
Sponsors may also be deterred by regulatory concerns. For instance, to be able to issue CAT bonds, sponsors need to establish an SPV or SPI. Regulatory frameworks should: enable the establishment of such an entity with appropriate and clear procedures and requirements; provide governing standards related to the management and administration of the SPV or SPI; and define the reporting system. Lack of SPV or SPI regulatory frameworks in developing countries is a major reason why CAT bonds are often issued offshore. Among Dynamic Asian countries, the Philippines and Thailand have issued regulations to govern the functioning of SPVs. However, the scope is limited to very specific transactions, such as the purchase of non-performing assets in the Philippines (BSP, 2002[42]) and securitisation transactions in Thailand (SEC, 1997[43]).
Information asymmetry and insufficient investor protection are also challenges that need to be addressed. Price transparency is essential for secondary market trading. Information asymmetry can deter investors from purchasing CAT bonds and thus hamper market liquidity. Investors may be reluctant to take on risks if the sponsor is assumed to have superior information. Using insights from behavioural economics, Froot (1999[44]) argues that the greater the information asymmetry, the greater the risk of adverse selection of transactions against the investor. This is because investors might be concerned that sponsors who consider the cost of protection low are those whose risk is greater than appreciated. This effect worsens as the cost of securing financial coverage increases. Relatedly, Li et al. (2019[45]) contend that greater disclosure of information means a lower transaction cost, which implies that the respective CAT bonds also display higher liquidity.
Issuers of CAT bonds are not requested to report on a regular basis, unlike issuers of traditional bonds. As a result, many of the investor protection rules common to most traditional registered investments are missing in the case of CAT bonds (FINRA, 2021[46]). Regulatory authorities therefore need to ensure that CAT bond sponsors disclose sufficient information on the state of the collateral securities and offer investors an ongoing view of the catastrophe risk.
Concerning the tax treatment of CAT bonds, the tax codes of many countries lack comprehensive guidance with regard to the clarity of the structure, the nature of the product and classification for tax purposes. This may hinder investors from engaging in transactions. Under International Financial Reporting Standards (IFRS), the accounting treatment of alternative risk transfer mechanisms depends on whether they are classified as reinsurance contracts or financial derivatives. Under IFRS, reinsurance accounting applies only to risk mitigation instruments that have an indemnity-based trigger (IFRS Foundation, 2021[47]).
This classification will determine the applicable tax regime. Due to the lack of specific guidance with respect to the treatment of CAT bonds, their classification for income tax purposes remains uncertain and highly dependent on the particular features of each tranche of each issuance. For instance, if CAT bonds are qualified as a financial instrument, whether they are classified as an asset or liability may affect the tax status of gains or losses. On the other hand, if CAT bonds are treated as a reinsurance contract, then CAT bondholders will be subject to the tax regime applicable to reinsurance products. However, because of their inability to be bifurcated into an equity and liability component, CAT bonds tend to be closer to a financial instrument than to reinsurance (Kaplan and Lefebvre, 2003[48]).
The 2020 amendments to existing accounting standards will nevertheless allow contracts that limit compensation to the settlement of the policyholder’s obligations to be classified as financial instruments under IFRS 9. The significant insurance risk included in the CAT bond contractual cash flows suggests that they would be accounted for at fair value through profit and loss (KPMG, 2020[49]). These developments are likely to bring more clarity to CAT bondholders. Policy makers in Dynamic Asia need to ensure that local accounting standards reflect these developments at the international level.
Policy directions for fostering catastrophe bond markets
Based on the benefits of CAT bonds for sovereign risk transfer, as well as the challenges associated with their adoption in Dynamic Asia and the Pacific, major policy recommendations can be drawn. These recommendations may serve as guidelines for government decision making regarding the development of new sovereign disaster risk management programmes or the enhancement of existing ones.
First, it is crucial to plan a grand design for sovereign risk transfer, focussing on the risk not covered by the private sector. CAT bonds are a key instrument for sovereign risk transfer and the reduction of protection gaps, and they should not be omitted. Building up know-how involves establishing expertise and experience regarding CAT bonds through training sessions and cross hirings and partnering with private firms and business schools. Developing tailor-made catastrophe risk models is important. Creating meteorological, hydrological and seismological services and investing in measurement infrastructure are also important. Moreover, data providers must be independent and have reliable processes plus trained personnel, and they need to fulfil high standards of data security. Minimising basis risk could be accomplished by establishing a risk pool with insurance portfolio to enable indemnity triggers, while maximising the correlation of parameter values and natural disaster losses if using a parametric trigger. Financing the CAT bonds spreads can be challenging for emerging market sovereigns. It is therefore important to plan early and integrate the costs for the whole disaster risk management programme into fiscal budget on an ongoing basis. In addition, it is important to broaden investor bases and capacity building needs to be strengthened further. Finally, developing the local currency bond market is critical.
Policy recommendations for fostering CAT bond markets include the following:
Formulate a grand design for disaster risk financing
Invest in measurement infrastructure
Enhance quality of data
Develop catastrophe risk models
Enhance capacity building
Broaden investor bases
Minimise basis risk
Prepare distribution schemes
Develop the local currency bond market
Formulating a grand design for disaster risk financing is key
Formulating a grand design from a long-term perspective is important for countries in Dynamic Asia and the Pacific, while recognising the importance of an integrated approach to disaster risk management and the contribution of risk assessment, risk awareness and risk prevention to the financial management of disaster risks. The OECD has recommendations that provides guidance for governments in building financial resilience to disaster risks (OECD, 2023[50]). They include the importance of:
i) promoting comprehensive disaster risks assessments to support the evaluation of potential financial impacts across the economy and population and allow for the identification of financial vulnerabilities and an assessment of the benefits of investments in risk reduction,
ii) supporting financial resilience of households, businesses, non-profit institutions and subnational governments to disaster risks and the availability and use of risk transfer and risk financing tools for disaster risks, which includes supporting initiatives to raise awareness of disaster risks,
iii) assessing and managing the financial impacts of disasters on public finances by evaluating the potential impacts of disasters on government, developing plans to ensure adequate funding, ensuring adequate plans are in place to disburse funds in a timely and equitable manner, building confidence in the government’s capacity to manage disaster risk financing, and assessing cost and benefit of risk retention, risk financing, or risk transfer,
iv) and establishing coherent strategies for building financial resilience to disasters that foster an integrated approach to managing the financial impacts of disaster risks across all levels of government, ensure sufficient institutional capacity and expertise for the implementation of these strategies, ensure co-operation and co-ordination across public and private sectors, leverage opportunities for international co-operation and information sharing considering the potential of cross-border impacts of disaster risks, and most importantly, take into account the characteristics, evolution and implications of different hazards.
Nevertheless, the direction of building up the disaster risk financing framework practically will differ by country. Broadly, there are two main pillars of function that policy makers need to consider in the grand design, namely risk pooling and risk transfer. Countries in the region need to strengthen both functions in parallel, though the way forward will be different depending on the level of each country’s development.
Pooling risk, typically in the form of insurance, improves resilience. Such pools may act either as an insurance carrier, offering policies for households and firms not available from the private sector, or as a reinsurer that enlarges the risk-bearing capacity of the country’s primary insurers. National insurance or reinsurance schemes have been used in OECD countries (Table 3.1) as well as in emerging markets. Examples include essentially public insurance carriers such as the National Flood Insurance Program (NFIP) in the United States and the Turkish Catastrophe Insurance Pool (TCIP) (Yazici, 2006[51]), as well as the French Caisse Centrale de Réassurance (CCR), which acts as a public-sector reinsurer. In parallel, governments should foster private insurance markets, so that private insurance companies could cover the main part of disaster losses suffered by households and firms. However, in Dynamic Asia and the Pacific, insurance products are often too expensive for many consumers, especially for low-income households, and many households lack a necessary knowledge of insurance. Insurers may thus find it hard to build a profitable business in most cases. The OECD (2021[52]) discusses several important points to develop disaster insurance programme. While establishing a catastrophe risk insurance programme to broaden insurance coverage, governments need to carefully consider the potential trade-offs inherent in different approaches to programme design, including:
Approaches designed to ensure coverage availability do not always result in broad coverage as policyholders may underestimate the risk of losses or have an expectation of government financial support should a large catastrophe occur and therefore not acquire the available insurance coverage.
Efforts to support affordability through cross-subsidisation between policyholders can blunt incentives for risk reduction and can raise issues of fairness if cross-subsidies benefit wealthier policyholders that could afford to pay higher premiums, although some mutualisation may be necessary for some risks to become insurable.
Subsidisation of the aggregate cost of programme coverage can put taxpayers at risk and might also raise competition concerns if the coverage provided by catastrophe risk insurance programmes competes directly with coverage provided by private (re)insurers.
Limiting the scope or amount of coverage provided by a catastrophe risk insurance programme to specific perils or policyholders can reduce public sector exposure although may lead to gaps in coverage and can also reduce the ability of the programme to benefit from diversification.
Catastrophe risk insurance programmes can play an important role in developing modelling and risk analytics tools – particularly for perils that have not traditionally created significant exposure for private (re)insurers – although limiting private sector involvement in the assumption of risk could hamper the development of private sector models and analytics.
Catastrophe risk insurance programmes can provide a source of expertise and funding to support risk reduction although their capacity to contribute will depend on the scope of the coverage that they provide (and the amount of premiums that they collect).
Careful consideration should also be given to the differences in the characteristics of the underinsured peril. By nature, some perils are more challenging to quantify or lead to high levels of correlation in losses:
Quantifying the financial consequences of infectious disease outbreaks, for example, involves uncertainties related to not only the frequency and severity of outbreaks, but also to the response of public authorities and individuals as well as the capacity of public health systems to manage the health impacts.
A number of perils (e.g. cyber risk, infectious disease outbreaks) can materialise as both low- and high-severity events with not all occurrences of the peril leading to catastrophic losses.
Perils also differ in terms of the level of correlation across countries and the diversification benefits that can be achieved in a global portfolio. Cyber risks and pandemics, for example, cannot necessarily be diversified by assuming risk in different countries.
All of these factors affect the ability of private insurance and reinsurance markets to assume risk. They will also require different approaches to the design of any catastrophe risk insurance programme.
To increase the financial preparedness of countries and thus improve their disaster risk resilience, another strategy is risk transfer, typically through market-based solutions such as insurance linked securities or catastrophe bonds. They should be an integral pillar of any integrated grand design of disaster risk management strategy. CAT bonds should be part of any diversified sovereign risk management strategy for emerging countries with major natural disaster exposure. As discussed above, sovereign risk transfer through CAT bonds has various advantages, such as a fully collateralised, flexible and immediate alternative for classical insurance coverage, and additionally guarantees multi-year coverage and price stability.
If a country formulates a grand design of disaster risk financing strategy, it is important to recognise development of capital and insurance market, potential differences in fiscal resources and repayment capacities, and other key factors that may influence financial strategies for disaster risk, such as data availability and technical expertise (OECD, 2022[2]).
Table 3.1. Catastrophe risk insurance programmes in selected OECD countries
Note: 1. There are public insurers for real estate in 19 Swiss cantons. The information provided in the table is for Gebäudeversicherung Graubündenin the canton of Grisons (as an illustrative example).
1. There are residual insurance arrangements that offer coverage for all or some property risks in many US states. Similar to Citizens in Florida, these programmes are aimed at making insurance coverage available to households that are unable to secure coverage in the private market.
Source: Authors’ adaptation of Table 2.1 from Enhancing Financial Protection Against Catastrophe Risks: The Role of Catastrophe Risk Insurance Programmes (OECD, 2021) (including table notes).
Investing in measurement infrastructure is important for data collection
The creation of national meteorological, hydrological, and seismological services akin to the US National Oceanic and Atmospheric Association (NOAA) and the United States Geological Survey (USGS) is important. These are government agencies responsible for collecting and analysing data on specific natural events such as cyclones, floods, wildfires or earthquakes. Such services play a critical role in understanding and managing the natural disaster risk that emerging countries in Asia and the Pacific may want to transfer to capital markets via sovereign CAT bonds.
Beyond this organisational prerequisite, governments need to stress test their existing data measurement, data transmission and data storage infrastructure. Many emerging countries already have measurement networks in place, such as meteorological and hydrological monitoring systems. These networks should be improved by investing in denser geographical coverage and more reliable and resilient devices. They also need reliable maintenance plans to ensure their functionality in the long run. If capital market investors doubt the accuracy and reliability of the measurement infrastructure, they will demand substantial spread markups or refrain from purchasing the sovereign CAT bonds altogether.
In addition to existing measurement stations, governments can consider the use of advanced monitoring technologies, such as remote sensing, satellite imagery and permanent drone surveillance. These may deliver real-time data on natural disasters and enable technological leapfrogging compared to classical measurements for certain perils (e.g. floods and drought). Satellite imagery in particular can be a powerful means of assessing losses, including the number of buildings affected and the severity of the damage. Apart from determining the CAT bond payout, this information can also be used by governments to allocate resources for relief and recovery.
The improvement of data measurement infrastructure for natural disaster risk will clearly be associated with substantial public investments in technology and people. It may therefore also require international partnerships or development aid. Governments of emerging countries that want to engage in sovereign risk transfer via CAT bonds should hence foster such partnerships with other countries and international organisations to access the required technical expertise and funding. There are several precedents that document how such partnerships can be fruitful. In 2008, for example, the Chinese central government got involved in a pilot for parametric insurance in Anhui province along with local insurance companies and international organisations (UN ESCAP, 2015[25]).
Enhancing quality of data is crucial
Accurate and timely data is critical for effective disaster risk transfer. As discussed above, emerging countries must improve the availability of data on natural disasters by investing in their data measurement and processing infrastructure. Policy makers in Dynamic Asia and the Pacific should develop databases at the national or regional level to track parametric data on the characteristics of various types of natural hazards. Enhanced parametric data on a wide range of catastrophe events will support the modelling of additional types of perils that are not covered by existing approaches. However, it will be just as important to establish trustworthy data providers. Trustworthy data providers deliver accurate, reliable and up-to-date information that can be used with confidence for decision-making purposes. To this end, they need standard operating procedures and personnel who are highly trained in all matters of data management. The data providers should also be transparent about their sources, methodologies and any limitations or caveats associated with the data.
In addition to these key aspects, the data providers should have appropriate measures in place to protect the data from unauthorised access or breaches. Data security is of critical importance in today’s interconnected world, where personal, financial and business information is shared and stored digitally. The increasing reliance on digital information has led to gigantic amounts of data being created and processed, which increases the danger of cyberattacks and unauthorised access.
Finally, the independence of the data provider from the government that acts as the CAT bond cedent is of utmost importance. Investors will not be prepared to invest in CAT bonds if they have reason to believe that the government is able to influence the underlying and provoke a payout of the CAT bond that would be unjustified based on objective reporting. A straightforward way to solve this problem is to use a modelled loss trigger and select one of the established catastrophe modelling firms (RMS, AIR Worldwide, CoreLogic) as calculation agent. For example, AIR Worldwide continuously adds new territories and perils to its models. The firm has already begun to expand its Southeast Asia earthquake and typhoon models to smaller countries such as Guam, Macau and the island of Saipan (Verisk, 2016[53]). AIR Worldwide illustrates the importance of an independent third party in non-indemnity CAT bond transactions: when the IBRD CAR 123‑124 CAT bond was threatened by Typhoon Noru and the Philippine government requested an event calculation process and AIR Worldwide modelled the loss and announced that the cyclone did not qualify as a trigger event (Artemis, 2022[54]; Evans, 2022[55]).
Developing catastrophe risk models is necessary for success
Owing to the infrequent nature of large-scale natural disasters, historical event data does not convey a complete picture of the parameter or loss distributions for CAT bonds (Brookes, 2009[24]). There are simply not enough observations to support the tail. Catastrophe risk models fill this data void by simulating a myriad of artificial events. Today, both advanced science and computing power are available to maintain accurate catastrophe risk models. Different modelling approaches for catastrophe risk pricing are discussed in Annex 3.A. Lane (2022[56]) compares the catastrophe loss experience between 2001 and 2020 to the expected loss estimates generated by catastrophe risk models and concludes that analysis provided by catastrophe modelling firms “gives an accurate characterisation of the risks embedded in the ILS they are considering acquiring”. The CAT bond market has come to accept these models a basis for pricing and risk management.
However, just as the main part of the CAT bond market mirrors the largest primary insurance markets around the world (see Chapter 2), so do the most accurate models maintained by catastrophe modelling firms. After all, these firms are for-profit organisations with commercial interests. Thus, the catastrophe risk modelling know-how and capacity available for emerging countries is smaller than for developed economies. Particularly for meteorological disasters, such as cyclones and droughts, the modelling capabilities in developing countries are less pronounced (White et al., 2022[57]). Regarding earthquakes, models for most middle- and high-income developing countries tend to be available, but often lack sufficient data on buildings and infrastructure, which impedes a proper estimation of economic losses (Mahul and Cummins, 2009[58]), which is particularly relevant for certain types of triggers such as modelled loss, industry loss index or indemnity.
The situation in Dynamic Asia and the Pacific is heterogeneous. Several private companies and organisations, such as RMS and AIR Worldwide, have developed catastrophe risk models for Asian countries (Mahul and Cummins, 2009[58]). To estimate the likelihood and potential impact of future catastrophes, these models use a combination of data on historical events, scientific understanding of the hazard, and information on the exposed assets and populations. For example, AIR Worldwide maintains typhoon, earthquake and crop risk models for China. However, the natural disaster risk in other parts of Asia is not yet sufficiently captured by catastrophe risk models, which often suffer from limited availability and quality of data and lack a proper regional specification (White et al., 2022[57]).
While catastrophe risk models have proved a valuable tool for risk assessment, pricing and management of CAT bonds, the application of these models in the Asian context is fraught with challenges. To address these issues, there is a pressing need for concerted efforts from both public and private entities to improve data availability and quality, refine modelling techniques and keep pace with the evolving risk landscape in Asia. This would help to enhance the reliability of catastrophe risk models, facilitating the use of CAT bonds as a tool for sovereign risk transfer in the region.
The successful usage of CAT bonds for sovereign risk transfer crucially depends on the availability of reliable catastrophe risk models. Many countries in Dynamic Asia and the Pacific are exposed to natural perils such as floods, cyclones and earthquakes (Swiss Re, 2022[1]). Before these risks can be transferred to capital markets, they must be modelled. Without proper risk quantification, pricing and risk transfer are not feasible. However, due to the low-frequency, high-severity character of the risks, historical data is not sufficient for risk quantification (Brookes, 2009[24]). Catastrophe models could fill the gap but existing ones are mainly available for established insurance markets, such as the United States or Europe.
Opening access to the CAT bond market to a broader range of investors would require public disclosure of prices, offerings and any other information necessary for investors to assess the risks associated with investing in this asset class. The increasing availability of data and computing power implies that data-driven models of risk pricing will be increasingly sought. Applications from machine learning have the potential to improve the performance of these models.
Sovereigns will also need to establish open data and research policies for the development of catastrophe risk models, making these inputs publicly available (e.g. through public repositories or online platforms) and ensuring that they can be accessed by researchers. Without the sharing of data and information on the underlying risk, such as historical weather and seismological data, risk modelling will be an ill-fated task. After all, catastrophe risk models must be capable of estimating the tail of the loss distribution, while also matching empirically observed loss experience (Moody's RMS, 2023[59]). This point is closely interlinked with the next two policy recommendations discussed below: investing in data measurement infrastructure and reliable data providers. In most emerging countries, the quality and detail of scientific data are not on the same standards as in the United States and Europe. Yet better data is a key condition for better models. Governments should therefore also engage in an ongoing effort to increase the amount and improve the precision of data on the natural perils to which their countries are exposed.
Enhancing capacity-building
The adoption of CAT bonds needs to be accompanied by a build-up of expertise and experience. CAT bonds are complex financial instruments that require cedents to understand both reinsurance and financial markets (Ben Ammar, Braun and Eling, 2015[60]). Experience and expertise are also critical factors for adoption on the investor side Emerging countries aiming to deploy CAT bonds therefore need to establish a dedicated unit or task force inside their public administration bodies.
At the same time, understanding financial competence in the context of CAT bonds to be something larger than management of personal finances is essential for the capacity-building of policy makers. Governments of OECD countries, Dynamic Asia and the Pacific offer financial training to their staff, but the training focuses on management of personal finances rather than skills necessary for policy makers in their official functions. Among the specific topics necessary to examine, policy makers need to understand not just the benefits of CAT bonds, but also their practical implications. A clear regulatory framework, especially regarding taxation, is essential both for policy makers and investors.
Increasing the capacity of policy makers to take advantage of CAT bonds requires a whole-of-government approach. Officials in ministries of economy, finance, and disaster management would benefit from such training depending on the governance and budgeting structures of a given country. Tying career advancement to upskilling would provide an incentive for employees to participate. Policy makers and tertiary educational institutions in Dynamic Asia and the Pacific should collaborate to inform each other of needs and trends, keeping the training current. Post-secondary institutions could also develop relevant courses for students to create a human capital pipeline.
One example of an existing training course on CAT bonds is “The Fundamentals of Insurance-Linked Securities (ILS)”, offered by a UK reinsurer specialising in catastrophe reinsurance (Phoenix CRetro, n.d.[61]). The online course lasts seven weeks and covers an explanation of why risk transfer is important; an introduction to ILS (of which CAT bonds are one type); pricing and legal considerations (that may also have utility for policy makers); advantages and disadvantages; case studies; and in-depth discussion of how ILS can provide disaster risk financing and meet environmental, social and governance (ESG) goals. The course is open to a wide variety of participants including advanced students and finance professionals or policy makers looking to upskill. Recognising the barriers faced by course participants from developing countries and the importance of the subject matter to their prosperity, the course offers tuition reduction for participants from developing countries (Phoenix CRetro, n.d.[61]).
Broadening investor bases
Training is key to broadening investor bases, something that should be a central goal for policy makers seeking to develop CAT bond markets. The collection of data on financial capability disaggregated over many demographics will be required for authorities to gain a fuller picture of who is currently excluded from capital market participation. In some countries, data collection will need to be conducted in multiple languages and the results aggregated where appropriate. Producing training material in multiple languages will necessarily require a certain amount of duplicated effort. While working from a common base, policy makers may consider the implementation of targeted training programmes that emphasise different concepts depending on the needs of each demographic.
In addition to training in the technical aspects of investing, training for overcoming other barriers such as low personal confidence, and lack of trust in experts is also essential. For instance, if investor bases are to be expanded to include more women, then women must receive training suited to their needs. Some women may have a lower risk appetite than many men and less confidence in their abilities due to factors largely beyond their control. Becoming a professional investor does not change this, so training programmes targeted to women should seek to boost their confidence. In addition, self-confidence must align with an individual’s level of financial ability and knowledge. For this alignment to take place, it is important that potential investors receive advice from trustworthy and knowledgeable sources. Unfortunately, in their quest for trustworthiness, people often turn to others who may also have low expertise, such as friends or family, which risks perpetuating financial mistakes (van Rooij, Lusardi and Alessie, 2011[62]). For both institutional and retail investors, authorities must foster trust in credentialed experts as superior sources of financial advice.
Minimising basis risk may require several steps
Basis risk means the possibility that the payout of a CAT bond will not perfectly match the actual catastrophe losses suffered by the cedent. It is a concern if the CAT bond relies on non-indemnity triggers. To tackle basis risk in CAT bonds, governments in the countries of Dynamic Asia and the Pacific may take several steps, the simplest of which would be to avoid non-indemnity triggers. However, this requires an insurance portfolio that can be referenced by the CAT bond transaction. In this regard, governments would additionally need to establish a national risk pool that underwrites a policy portfolio from households and businesses. A major challenge here is the distribution of insurance policies to households and firms. In developed countries, national risk pools can tap into the distribution networks of the private insurance sector. An example is the NFIP, which sells policies through insurance agents and brokers. This will be difficult in some countries of Dynamic Asia and the Pacific, such as Viet Nam, because traditional insurance distribution channels are less established (KPMG, 2022[63]). However, digital-direct sales and bancassurance constitute viable alternatives (Gonulal, Goulder and Lester, 2012[64]).
When parametric triggers are chosen or deemed necessary, the minimisation of basis risk is clearly dependent on the development and testing of sophisticated catastrophe risk models and the availability of a reliable measurement infrastructure. The policy recommendations discussed above are thus fundamental prerequisites for the minimisation of basis risk, too. Given the output of a proper catastrophe risk model, it will be possible to quantify basis risk (Brookes, 2009[24]). Subsequently, the key characteristics of the CAT bond (e.g. the geography and layer) can be modified to maximise the stochastic dependence of the physical trigger parameter (e.g. wind speed) with the disaster losses that are expected under any given catastrophe scenario. In doing so, sponsors can minimise the expected shortfall of the payout across all trigger scenarios.
Finally, the choice of parameters should be aligned with exposure in the best possible way. This can be achieved by switching from pure parametric to parametric index triggers. The latter allow cedents to apply a weighting to the readings from different measurement stations that best mirrors their actual exposure (ADB, 2009[65]). The weighted sum of parameter values then constitutes the parametric index. To build a parametric index that is most correlated with losses, accurate historical data and, again, advanced modelling capabilities are imperative.
Preparing distribution schemes ex ante will help to avoid delays in relief and recovery
To ensure rapid and targeted distribution of the proceeds from sovereign CAT bonds in the event of a disaster, contingency plans (protocols and guidelines for disaster responses) must be put in place ex ante. Slow political processes may otherwise substantially reduce the effectiveness of CAT bonds for immediate relief and a timely recovery of essential public infrastructure. In emerging countries with a weak healthcare system and disconnected rural areas that quickly run out of supplies, every hour of delay adds to the suffering of the affected population.
Emrich, Aksha and Zhou (2022[35]) recommend that social vulnerabilities be considered for the allocation of disaster assistance across local governments, households and businesses. Hence, the distribution of the CAT bond proceeds in a contingency plan should be tied to a detailed needs assessment that reflects both the extent of the damage and the financial resilience of the affected communities.
Moreover, the contingency plans should determine how the CAT bond proceeds will be used. This includes transparent accounting and monitoring mechanisms. Establishing appropriate rules ex ante ensures that those in charge of the distribution cannot easily misappropriate funds. Immediate relief can take different forms, such as cash, vouchers or in-kind assistance. Direct cash transfers enable households and businesses to decide for themselves how they would like to achieve the best possible improvement of their situation. Cash is considered to be more effective in disasters than in-kind support through medical supplies, potable water, food, etc.
Developing the local-currency bond market
Local-currency government bond markets provide the necessary platform and institutional framework for the issuance of catastrophe bonds in the region. The trust both local and international investors have in local-currency-denominated government bonds fosters confidence in catastrophe bonds. This is expected to lead to increased local and foreign investor participation.
Several other studies have been conducted to determine the key factors for the development of local-currency bond markets (LCBM). Claessens et al. (2007[66]) find that measures to expand investor bases matter, such that economies with larger domestic financial systems in terms of bank deposits and stock market capitalisation have deeper local currency bond markets. Essers et al. (2015[67]) conclude that lower fiscal balances and inflation, along with higher institutional quality, are associated with larger LCBM capitalisation. Berensmann, Dafe and Volz (2015[68]) highlight the importance of foreign investor participation in LCBM and state that it enhances market size by broadening the investor base. Eichengreen and Luengnaruemitchai (2004[69]) show that major factors contributing to the development of national bond markets are the size of an economy, strong institutional structure and more stable exchange rates. Bhattacharyay (2013[70]) finds that, while the size of the economy associates positively with bond market deepening in both sectors, the size of the banking sector correlates positively with government bond market capitalisation but negatively with corporate-sector bond market depth. Molnar-Tanaka and Imisiker (2023[71]) show that elements affecting LCBM depth include macroeconomic factors such as GDP, inflation and fiscal balance, as well as the exchange-rate regime, capital account openness, and creditor rights. They find that depth of financial markets and institutions, and access to them, have significantly positive relationships with LCBM development.
Overall, a solid macroeconomic framework, as well as a solid institutional framework have been well-established as the key components of the foundation for a robust LCBM. Development of local currency-denominated catastrophe bonds in Asia and the Pacific should occur in conjunction with the development of local currency government bond markets more broadly, though there is no universal approach.
Conclusion
This chapter has discussed catastrophe bonds as a potential sovereign risk transfer tool for Dynamic Asia and the Pacific. It has highlighted both the benefits of CAT bonds for sovereign risk transfer in countries in the region and the major challenges that need to be overcome to realise these benefits. Finally, the chapter has offered policy recommendations for fostering CAT bond markets in the region.
The benefits of using CAT bonds include: their role as an alternative to private insurance; diversification of coverage; flexible multi-year coverage; full collateralisation; transparency and fast settlement; efficient price discovery; the non-peak territory discount; and multi-country cost-sharing and spread discount. Policy recommendations for developing CAT bond markets include preparing a grand design for disaster risk financing; enhancing capacity building; broadening investor bases; developing catastrophe risk models; investing in measurement infrastructure; establishing trustworthy data; minimising basis risk; preparing distribution schemes and; developing local-currency bond markets.
References
[65] ADB (2009), Natural Catastrophe Risk Insurance Mechanisms for Asia and the Pacific: Main Report, Asian Development Bank, Manila, https://www.adb.org/sites/default/files/publication/27991/natural-catastrophe-risk-insurance.pdf.
[26] AFC (2011), Report on Impact Evaluation of Pilot Weather Based Crop Insurance Study (WBCIS), Government of India, https://pmfby.gov.in/compendium/General/2011%20-%20Report%20on%20Impact%20Evaluation%20of%20Pilot%20Weather%20Based%20Crop%20Insurance%20Study%20(WBCIS).pdf.
[97] Agriculture Times (2021), Gramcover launches parametric insurance to protect farmers against weather vagaries, https://agritimes.co.in/farmers/gramcover-launches-parametric-insurance-to-protect-farmers-against-weather-vagaries/.
[39] Artemis (2022), Catastrophe Bond & Insurance-Linked Securities Deal Directory, https://www.artemis.bm/deal-directory/.
[54] Artemis (2022), Super typhoon Noru did not trigger Philippines catastrophe bond, https://www.artemis.bm/news/super-typhoon-noru-did-not-trigger-philippines-catastrophe-bond/.
[11] Artemis (2017), Mexico confirms $150m cat bond payout for quake, https://www.artemis.bm/news/mexico-confirms-150m-cat-bond-payout-for-quake/ (accessed on 31 January 2022).
[21] Artemis (2014), No Odile loss for MultiCat Mexico 2012 catastrophe bond, https://www.artemis.bm/news/no-odile-loss-for-multicat-mexico-2012-catastrophe-bond/.
[22] Artemis (2012), Caribbean Catastrophe Risk Insurance Facility: No payouts from Sandy, https://www.artemis.bm/news/caribbean-catastrophe-risk-insurance-facility-no-payouts-from-sandy/.
[40] Bantwal, V. and H. Kunreuther (2000), “A Cat Bond Premium Puzzle?”, Journal of Psychology and Financial Markets, Vol. 1/1, pp. 76-91, https://doi.org/10.1207/s15327760jpfm0101_07.
[24] Barrieu, P. and L. Albertini (eds.) (2009), Risk Modelling and the Role and Benefits of Cat Indices, John Wiley & Sons.
[74] Baryshnikov, Y., A. Mayo and D. Taylor (1998), “Pricing of CAT Bonds”, Working Paper, Dept. of Mathematics, University of Osnabrück.
[14] Beer, S. and A. Braun (2022), “Market-consistent valuation of natural catastrophe risk”, Journal of Banking & Finance, Vol. 134, p. 106350, https://doi.org/10.1016/j.jbankfin.2021.106350.
[60] Ben Ammar, Braun and Eling (2015), “Alternative Risk Transfer and Insurance-Linked Securities: Trends, Challenges and New Market Opportunities.”, I.VW HSG Schriftenreihe No. 56, Verlag Institut für Versicherungswirtschaft der Universität St. Gallen, St. Gallen, https://www.econstor.eu/handle/10419/226642.
[68] Berensmann, K., F. Dafe and U. Volz (2015), “Developing local currency bond markets for long-term development financing in Sub-Saharan Africa”, Oxford Review of Economic Policy, Vol. 31/3-4, pp. 350-378, https://doi.org/10.1093/oxrep/grv032.
[70] Bhattacharyay, B. (2013), “Determinants of bond market development in Asia”, Journal of Asian Economics, Vol. 24, pp. 124-137, https://doi.org/10.1016/j.asieco.2012.11.002.
[13] Blackman, J., M. Maidenberg and S. O’Regan (2018), Mexico’s disaster bonds were meant to provide quick cash after hurricanes and earthquakes. But it often hasn’t worked out that way, https://www.latimes.com/world/mexico-americas/la-na-mexico-catastrophe-bonds-20180405-htmlstory.html.
[19] Braun, A. (2015), “Pricing in the Primary Market for Cat Bonds: New Empirical Evidence”, Journal of Risk and Insurance, Vol. 83/4, pp. 811-847, https://doi.org/10.1111/jori.12067.
[38] Braun, A., K. Müller and H. Schmeiser (2013), “What Drives Insurers’ Demand for Cat Bond Investments? Evidence from a Pan-European Survey”, The Geneva Papers on Risk and Insurance - Issues and Practice, Vol. 38/3, pp. 580-611, https://doi.org/10.1057/gpp.2012.51.
[42] BSP (2002), The Implementing Rules and Regulations of the Special Purpose Vehicle (SPV) Act of 2002, https://www.bsp.gov.ph/Regulations/Banking%20Laws/SPAV_IRR.pdf.
[75] Burnecki, K. and G. Kukla (2003), “Pricing of zero-coupon and coupon cat bonds”, Applicationes Mathematicae, Vol. 30/3, pp. 315-324, https://doi.org/10.4064/am30-3-6.
[76] Burnecki, K., G. Kukla and D. Taylor (2005), “Pricing of Catastrophe Bonds”, Statistical Tools for Finance and Insurance, pp. 93-114, https://link.springer.com/content/pdf/10.1007/3-540-27395-6_4.pdf.
[89] Canabarro, E. et al. (2000), “Analyzing Insurance‐Linked Securities”, The Journal of Risk Finance, Vol. 1/2, pp. 49-75, https://doi.org/10.1108/eb043445.
[87] Carayannopoulos, P. and M. Perez (2014), “Diversification through Catastrophe Bonds: Lessons from the Subprime Financial Crisis”, The Geneva Papers on Risk and Insurance - Issues and Practice, Vol. 40/1, pp. 1-28, https://doi.org/10.1057/gpp.2014.14.
[30] Chatoro, M. et al. (2023), “Catastrophe bond pricing in the primary market: The issuer effect and pricing factors”, International Review of Financial Analysis, Vol. 85, p. 102431, https://doi.org/10.1016/j.irfa.2022.102431.
[66] Claessens, S., D. Klingebiel and S. Schmukler (2007), “Government Bonds in Domestic and Foreign Currency: the Role of Institutional and Macroeconomic Factors*”, Review of International Economics, Vol. 15/2, pp. 370-413, https://doi.org/10.1111/j.1467-9396.2007.00682.x.
[4] Clarke, D. et al. (2017), “Evaluating Sovereign Disaster Risk Finance Strategies: A Framework”, The Geneva Papers on Risk and Insurance - Issues and Practice, Vol. 42/4, pp. 565-584, https://doi.org/10.1057/s41288-017-0064-1.
[81] Cox, J., J. Ingersoll and S. Ross (1985), “A Theory of the Term Structure of Interest Rates”, Econometrica, Vol. 53/2, pp. 385-407, https://pages.stern.nyu.edu/~dbackus/BCZ/discrete_time/CIR_Econometrica_85.pdf.
[72] Cox, S. and H. Pedersen (2000), “Catastrophe Risk Bonds”, North American Actuarial Journal, Vol. 4/4, pp. 56-82, https://doi.org/10.1080/10920277.2000.10595938.
[41] Cummins, J. (2005), “Convergence in Wholesale Financial Services: Reinsurance and Investment Banking”, The Geneva Papers on Risk and Insurance - Issues and Practice, Vol. 30/2, pp. 187-222, https://doi.org/10.1057/palgrave.gpp.2510031.
[6] Cummins, J. and P. Trainar (2009), “Securitization, Insurance, and Reinsurance”, Journal of Risk and Insurance, Vol. 76/3, pp. 463-492, https://doi.org/10.1111/j.1539-6975.2009.01319.x.
[31] Dieckmann, S. (2010), “By Force of Nature: Explaining the Yield Spread on Catastrophe Bonds”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.1082879.
[36] Drakes, O. et al. (2021), “Social vulnerability and short-term disaster assistance in the United States”, International Journal of Disaster Risk Reduction, Vol. 53, p. 102010, https://doi.org/10.1016/j.ijdrr.2020.102010.
[69] Eichengreen, B. and P. Luengnaruemitchai (2004), “Why Doesn’t Asia Have Bigger Bond Markets?”, NBER Working Paper No. w10576, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=559226.
[35] Emrich, C., S. Aksha and Y. Zhou (2022), “Assessing distributive inequities in FEMA’s Disaster recovery assistance fund allocation”, International Journal of Disaster Risk Reduction, Vol. 74, p. 102855, https://doi.org/10.1016/j.ijdrr.2022.102855.
[67] Essers, D. et al. (2015), “Local Currency Bond Market Development in Sub-Saharan Africa: A Stock-Taking Exercise and Analysis of Key Drivers”, Emerging Markets Finance and Trade, Vol. 52/5, pp. 1167-1194, https://doi.org/10.1080/1540496x.2015.1073987.
[55] Evans, S. (2022), Philippines requests cat bond event calculation for super typhoon Noru, https://www.artemis.bm/news/philippines-cat-bond-event-calculation-super-typhoon-noru/.
[46] FINRA (2021), Insurance-Linked Securities, https://www.finra.org/investors/insights/insurance-linked-securities.
[44] Froot, K. (1999), “The Evolving Market for Catastrophic Event Risk”, Risk Management and Insurance Review, Vol. 2/3, pp. 1-28, https://doi.org/10.1111/j.1540-6296.1999.tb00001.x.
[86] Galeotti, M., M. Gürtler and C. Winkelvos (2012), “Accuracy of Premium Calculation Models for CAT Bonds—An Empirical Analysis”, Journal of Risk and Insurance, Vol. 80/2, pp. 401-421, https://doi.org/10.1111/j.1539-6975.2012.01482.x.
[5] Gallagher Re (2022), Gallagher Re 1st view: Changing environment, https://www.ajg.com/gallagherre/news-and-insights/2022/july/gallagher-re-1st-view-1-july-2022/.
[64] Gonulal, S., N. Goulder and R. Lester (2012), Bancassurance-A Valuable Tool for Developing Insurance in Emerging Markets, The World Bank, https://doi.org/10.1596/1813-9450-6196.
[15] Götze, T. and M. Gürtler (2020), “Risk transfer and moral hazard: An examination on the market for insurance-linked securities”, Journal of Economic Behavior & Organization, Vol. 180, pp. 758-777, https://doi.org/10.1016/j.jebo.2019.06.010.
[79] Härdle, W. and B. Cabrera (2010), “Calibrating CAT Bonds for Mexican Earthquakes”, Journal of Risk and Insurance, Vol. 77/3, pp. 625-650, https://doi.org/10.1111/j.1539-6975.2010.01355.x.
[90] Heaton, J. and D. Lucas (1996), “Evaluating the Effects of Incomplete Markets on Risk Sharing and Asset Pricing”, Journal of Political Economy, Vol. 104/3, pp. 443-487, https://www.jstor.org/stable/2138860.
[18] Herrmann, M. and M. Hibbeln (2022), “Trading and liquidity in the catastrophe bond market”, Journal of Risk and Insurance, Vol. 90/2, pp. 283-328, https://doi.org/10.1111/jori.12407.
[47] IFRS Foundation (2021), IFRS 4 insurance contracts, https://www.ifrs.org/issued-standards/list-of-standards/ifrs-4-insurance-contracts/#about.
[78] Jarrow, R. (2010), “A simple robust model for Cat bond valuation”, Finance Research Letters, Vol. 7/2, pp. 72-79, https://doi.org/10.1016/j.frl.2010.02.005.
[48] Kaplan, S. and G. Lefebvre (2003), “CAT bonds: Tax treatment of an innovative financial product”, Journal of Taxation of Financial Institutions, Vol. 16/4, https://www.artemis.bm/articles/Kaplan2.pdf.
[92] Karolyi, A. (ed.) (2020), “Empirical Asset Pricing via Machine Learning”, The Review of Financial Studies, Vol. 33/5, pp. 2223-2273, https://doi.org/10.1093/rfs/hhaa009.
[96] Kelly, B., S. Pruitt and Y. Su (2019), “Characteristics are covariances: A unified model of risk and return”, Journal of Financial Economics, Vol. 134/3, pp. 501-524, https://doi.org/10.1016/j.jfineco.2019.05.001.
[88] Kelman, I. (2001), “The autumn 2000 floods in England and flood management”, Weather, Vol. 56, pp. 346-360, https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/j.1477-8696.2001.tb06507.x.
[63] KPMG (2022), Asia Pacific Insurance Sector Opportunities: Navigating the Region’s Life and Non-Life M&A Landscape, https://assets.kpmg.com/content/dam/kpmg/cn/pdf/en/2022/12/asia-pacific-insurance-sector-opportunities.pdf.
[49] KPMG (2020), Insurance Contracts: First Impressions: 2020 Edition, https://assets.kpmg.com/content/dam/kpmg/ie/pdf/2020/09/ie-ifrs-17-first-impressions.pdf.
[8] Lakdawalla, D. and G. Zanjani (2011), “Catastrophe Bonds, Reinsurance, and the Optimal Collateralization of Risk Transfer”, Journal of Risk and Insurance, Vol. 79/2, pp. 449-476, https://doi.org/10.1111/j.1539-6975.2011.01425.x.
[56] Lane, M. (2022), “The ILS loss experience: natural catastrophe issues 2001–2020”, The Geneva Papers on Risk and Insurance - Issues and Practice, Vol. 49/1, pp. 97-137, https://doi.org/10.1057/s41288-022-00275-5.
[83] Lane, M. (2000), “Pricing Risk Transfer Transactions”, ASTIN Bulletin, Vol. 30/2, pp. 259-293, https://www.casact.org/sites/default/files/2021-02/2001-lane.pdf.
[17] Lane, M. and R. Beckwick (2016), TRACE data twenty one months on: ILS trade or quote data?, http://www.lanefinancialllc.com/content/view/361/67/.
[84] Lane, M. and O. Mahul (2008), Catastrophe Risk Pricing, World Bank, https://openknowledge.worldbank.org/server/api/core/bitstreams/b79389c4-b1e7-54dd-9cc4-dde0202db437/content.
[45] Li, X. et al. (2019), “Founders and the decision of Chinese dual-class IPOs in the U.S.”, Pacific-Basin Finance Journal, Vol. 57, p. 101017, https://doi.org/10.1016/j.pacfin.2018.04.009.
[58] Mahul, O. and J. Cummins (2009), Catastrophe Risk Financing in Developing Countries: Principles for Public Intervention, https://www.worldbank.org/en/programs/disaster-risk-financing-and-insurance-program/publication/catastrophe-risk-financing-in-developing-countries.
[16] Makariou, D., P. Barrieu and Y. Chen (2021), “A random forest based approach for predicting spreads in the primary catastrophe bond market”, Insurance: Mathematics and Economics, Vol. 101, pp. 140-162, https://doi.org/10.1016/j.insmatheco.2021.07.003.
[28] Matheswaran, K. et al. (2018), “Flood risk assessment in South Asia to prioritize flood index insurance applications in Bihar, India”, Geomatics, Natural Hazards and Risk, Vol. 10/1, pp. 26-48, https://doi.org/10.1080/19475705.2018.1500495.
[85] Mevorach, C. (2018), “Modelling Catastrophe Bond Pricing in the Primary Market - A Loglinear Approach”, University of Rochester, Rochester, New York, https://www.sas.rochester.edu/eco/undergraduate/papers/mevorach---catastrophe-bond-pricing,-2018.pdf.
[20] Michel-Kerjan, E. et al. (2011), “Catastrophe Financing for Governments: Learning from the 2009-2012 MultiCat Program in Mexico”, OECD Working Papers on Finance, Insurance and Private Pensions, No. 9, OECD Publishing, Paris, https://doi.org/10.1787/5kgcjf7wkvhb-en.
[23] Mobarak, A. and M. Rosenzweig (2013), “Informal Risk Sharing, Index Insurance, and Risk Taking in Developing Countries”, American Economic Review, Vol. 103/3, pp. 375-380, https://doi.org/10.1257/aer.103.3.375.
[71] Molnar-Tanaka, K. and S. Imisiker (2023), “Developing Local-Currency Bond Markets in Emerging Asia: Critical Factors, Challenges and Policy Actions”, Journal of Southeast Asian Economies, Vol. 40/3, pp. 318-343.
[59] Moody’s RMS (2023), Understanding catastrophe (CAT) modelling, https://www.rms.com/catastrophe-modeling?contact-us=cat-modeling.
[93] Mullainathan, S. and J. Spiess (2017), “Machine Learning: An Applied Econometric Approach”, Journal of Economic Perspectives, Vol. 31/2, pp. 87-106, https://doi.org/10.1257/jep.31.2.87.
[80] Nowak, P. and M. Romaniuk (2016), “Valuing catastrophe bonds involving correlation and CIR interest rate model”, Computational and Applied Mathematics, Vol. 37/1, pp. 365-394, https://doi.org/10.1007/s40314-016-0348-2.
[50] OECD (2023), Recommendation of the Council on Building Financial Resilience to Disaster Risks, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0436.
[2] OECD (2022), Building Financial Resilience to Climate Impacts: A Framework for Governments to Manage the Risks of Losses and Damages, OECD Publishing, Paris, https://doi.org/10.1787/9e2e1412-en.
[52] OECD (2021), Enhancing Financial Protection Against Catastrophe Risks: The Role of Catastrophe Risk Insurance Programmes, https://reliefweb.int/report/world/enhancing-financial-protection-against-catastrophe-risks-role-catastrophe-risk.
[34] O’Keefe, P., K. Westgate and B. Wisner (1976), “Taking the naturalness out of natural disasters”, Nature, Vol. 260/5552, pp. 566-567, https://doi.org/10.1038/260566a0.
[32] Papachristou, D. (2011), “Statistiscal Analysis of the Spreads of Catastrophe Bonds at the Time of Issue”, ASTIN Bulletin, Vol. 41(1), pp. 251-277, https://doi.org/10.2143/AST.41.1.2084394.
[10] Park, S. and X. Xie (2014), “Reinsurance and Systemic Risk: The Impact of Reinsurer Downgrading on Property–Casualty Insurers”, Journal of Risk and Insurance, Vol. 81/3, pp. 587-622, https://doi.org/10.1111/jori.12045.
[9] Park, S., X. Xie and P. Rui (2018), “The Sensitivity of Reinsurance Demand to Counterparty Risk: Evidence From the U.S. Property–Liability Insurance Industry”, Journal of Risk and Insurance, Vol. 86/4, pp. 915-946, https://doi.org/10.1111/jori.12244.
[61] Phoenix CRetro (n.d.), The Fundamentals of Insurance-Linked Securities (ILS), https://ils-course.com/.
[94] Rapach, D., J. Strauss and G. Zhou (2013), “International Stock Return Predictability: What Is the Role of the United States?”, The Journal of Finance, Vol. 68/4, pp. 1633-1662, https://doi.org/10.1111/jofi.12041.
[73] Reshetar, G. (2008), “Pricing of Multiple-Event Coupon Paying CAT Bond”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.1059021.
[7] SCOR (2022), SCOR successfully sponsors a new catastrophe bond, Atlas Capital Reinsurance 2022 DAC, https://www.scor.com/en/press-release/scor-successfully-sponsors-new-catastrophe-bond-atlas-capital-reinsurance-2022-dac.
[43] SEC (1997), Royal Enactment on Special Purpose Juristic Persons for Securitization B.E. 2540 (1997), https://www.sec.or.th/EN/Documents/EnforcementIntroduction/SPVen_codified.pdf.
[33] Spry, J. (2012), Non‐life Insurance Securitization: Market Overview, Background and Evolution, Wiley, https://doi.org/10.1002/9781119206545.ch2.
[95] Stambaugh, R. and Y. Yuan (2016), “Mispricing Factors”, The Review of Financial Studies, Vol. 30/4, pp. 1270-1315, https://doi.org/10.1093/rfs/hhw107.
[1] Swiss Re (2022), Natural catastrophes in 2021: the floodgates are open, https://www.swissre.com/institute/research/sigma-research/sigma-2022-01.html.
[3] Swiss Re (2018), ILS market update: August 2018, https://www.swissre.com/our-business/alternative-capital-partners/ils-market-update-aug-2018.html.
[82] Trottier, D. and V. Lai (2017), “Reinsurance or CAT Bond? How to Optimally Combine Both”, The Journal of Fixed Income, Vol. 27/2, pp. 65-87, https://doi.org/10.3905/jfi.2017.27.2.065.
[91] UN ESCAP (2019), “Chapter 1: The Asia-Pacific disaster riskscape”, in The Disaster Riskscape Across Asia-Pacific: Pathways for resilience, inclusion and empowerment. Asia-Pacific Disaster Report 2019, https://www.unescap.org/sites/default/d8files/APDR%202019%20Chapter%201.pdf.
[25] UN ESCAP (2015), Financing Disaster Risk Reduction for sustainable development in Asia and the Pacific, https://www.unescap.org/publications/financing-disaster-risk-reduction-sustainable-development-asia-and-pacific#.
[62] van Rooij, M., A. Lusardi and R. Alessie (2011), “Financial literacy and stock market participation”, Journal of Financial Economics, Vol. 101/2, pp. 449-472, https://doi.org/10.1016/j.jfineco.2011.03.006.
[77] Vaugirard, V. (2003), “Pricing catastrophe bonds by an arbitrage approach”, The Quarterly Review of Economics and Finance, Vol. 43/1, pp. 119-132, https://ideas.repec.org/a/eee/quaeco/v43y2003i1p119-132.html.
[53] Verisk (2016), AIR Worldwide Significantly Expands Its Model Coverage for Southeast Asia, https://www.verisk.com/newsroom/air-worldwide-significantly-expands-its-model-coverage-for-southeast-asia/.
[29] Whitehurst, D. et al. (2022), “Post-Flood Analysis for Damage and Restoration Assessment Using Drone Imagery”, Remote Sensing, Vol. 14/19, p. 4952, https://doi.org/10.3390/rs14194952.
[57] White, S. et al. (2022), The G7 takes on climate change: Are catastrophe bonds an answer?, https://www.milliman.com/en/insight/meeting-the-g7-commitment-to-disaster-financing-with-catastrophe-bonds#8.
[12] World Bank (2017), The impacts of insurance payouts on poverty: Estimating the effects of index-based livestock insurance in Mongolia, https://documents1.worldbank.org/curated/en/571231485212883943/pdf/ICRR-Disclosable-P088816-01-23-2017-1485212866628.pdf.
[27] World Food Programme (2010), The Potential for Scale and Sustainability in Weather Index Insurance for Agricultural and Rural Livelihoods, https://www.wfp.org/publications/potential-scale-and-sustainability-weather-index-insurance-agriculture-and-rural-livelihoods.
[37] Yago, G. and P. Reiter (2008), Financial Innovations Lab for Catastrophic Risk: Cat Bonds and Beyond, https://milkeninstitute.org/report/catastrophic-risk-cat-bonds-and-beyond.
[51] Yazici, S. (2006), “The Turkish Catastrophe Insurance Pool TCIP and Compulsory Earthquake Insurance Scheme”, in Catastrophic Risks and Insurance, OECD Publishing, Paris, https://doi.org/10.1787/9789264009950-20-en.
Annex 3.A. Modelling approaches and challenges for catastrophe risk pricing
Overview of CAT bond modelling approaches
The theoretical framework of CAT bond pricing was developed in particular after 1990 and is defined according to several parameters, notably location, period of coverage and disaster types (i.e. have defined triggers). For instance, if a qualifying disaster does not occur over the life of the bond, bondholders receive the face value at maturity. Conversely, if a qualifying disaster does occur, insurance companies are paid out of what bondholders would have otherwise received, and thus bondholders receive the face value less an amount determined by the payoff function of the bond. Payouts to insurance companies may be triggered by indemnity (i.e. the value of the sponsoring insurer’s actual losses), an index of industry-wide losses, measurable disaster parameters, such as earthquake magnitudes or wind speeds, modelled losses, or a hybrid of these. In any case, the relevant criteria must exceed defined thresholds for the payout to take place.
Stochastic models with discrete time
The literature on the valuation of CAT bonds is relatively sparse. In the specialised literature on CAT bond pricing, several authors apply stochastic models. Among these models, two advanced approaches that rely on stochastic processes with discrete time are those proposed by Cox and Pedersen (2000[72]) and Reshetar (2008[73]). The approach by Cox and Pedersen (2000[72]) uses the framework of the representative agent equilibrium, while Reshetar (2008[73]) assumes that the payout functions depend on catastrophic property losses and catastrophic mortality.
In the framework proposed by Cox and Pedersen (2000[72]), the CAT bond has a face value of 1, scheduled coupon payments c at the end of each period, and a final principal repayment of 1 at the end of the last period (T), if the catastrophe does not occur. Investors make an initial principal investment of 1. If the triggering event occurs, the payoff is a fraction of coupon and face value, which is f(1+c), denoting fraction with f. The payment is made in the end of the period during which the event occurs, and the bond is terminated. The general formula for the price at time 0 of the cash flow stream is as follows:
Where is the discount factor, price at time 0 of a zero-coupon bond maturing at time k, with face value 1. is the probability under risk-neutral measure that the catastrophe does not occur within the first k periods.
Next, the binomial term-structure model and binomial catastrophe risk structure are combined. is modelled as the catastrophe risk, and interest rate along with risk neutral probabilities of being in such state, which reflects in zero-coupon bonds . In the first period, there are four possible states:
i. interest rate goes up, catastrophe occurs, or
ii. interest rate goes up, catastrophe does not occur, or
iii. interest rate goes down, catastrophe occurs, or
iv. interest rate goes down, catastrophe does not occur.
The states are similar for the next periods. Once given the binomial model structure of rates along with risk neutral probabilities, and risk structure, meaning the catastrophe occurs or not, can be computed in each state. The above cash flow pricing formula is applied to determine the value of the CAT bond.
The framework proposed by Reshetar (2008[73]) implements theoretical pricing of a multiple-event CAT bond in an incomplete market setting using a representative agent pricing model. This model assumes that the belief of the representative agent corresponds to the weighted average of agents’ beliefs, while the dispersion in agents’ beliefs is incorporated into the discount rate of the representative agent.
Stochastic models with continuous time
The literature on stochastic models with continuous time for CAT bond pricing is comparatively broader than the one addressing stochastic models with discrete time. In an early attempt, Baryshnikov, Mayo and Taylor (1998[74]) use compound Poisson models to incorporate various characteristics of the CAT bond process. However, no analytical formula is derived in this approach. Burnecki and Kukla (2003[75]) correct the method proposed by Baryshnikov, Mayo and Taylor (1998[74]) to calculate non-arbitrage prices of a zero-coupon and coupon CAT bond.
Burnecki and Kukla (2003[75]) consider a bond with the payment of a certain amount Z at maturity time T contingent on threshold time τ > T, which is in fact a zero-coupon CAT bond (hereafter “ZCCB”). The condition required is that the payout processis a predictable process, which can be interpreted to mean that the payment at maturity is not directly linked to the occurrence and timing of the threshold. The amount Z can be, for instance, the principal plus interest, which is usually defined as a fixed percentage over the London Interbank Offered Rate (LIBOR).
The same approach is used in Burnecki, Kukla and Taylor (2005[76]). The underlying assumption in this model is that there is a Poisson point process of potentially catastrophe events. However, these events may or may not result in economic losses. The authors also assume that the economic losses associated with each of the potentially catastrophic events are independent and have a certain common probability distribution. Within this model, the threshold time can be seen as a point of a Poisson point process with a stochastic intensity depending on the instantaneous index position.
In a different approach, Vaugirard (2003[77]) develops an arbitrage method for pricing CAT bonds, which accounts for catastrophe events, interest rate randomness and non-traded underlying state variables. In the model, the bondholder expects to lose interest or a fraction of the principal if a natural risk index, whose value at date t is denoted It, hits a pre-specified threshold K. More specifically, if the index does not reach the threshold during a risk-exposure period T, the bondholder is paid the face value F. Otherwise, he receives the face value minus a write-down coefficient in percentage w. The author allows bond maturity T′ to be longer than the risk-exposure period T to account for possible lags in the risk-index assessment at expiration. IB(t) is the price of a ZCCB at time t, and TI,K is the first passage time of I through K.
Another approach is proposed by Jarrow (2010[78]), who obtains an analytically-closed CAT bond valuation formula, considering the LIBOR term structure of interest rates. The characteristics of the CAT bond value are as follows: the CAT bond receives floating payments based on the -year LIBOR rate Lt, paid in arrears, plus a spread . The face value of the bond, denoted A, is due at the bond’s maturity date, time T, unless there is a catastrophe event.
The CAT bond value is seen to be equal to:
i. The value of the next coupon payment, which equals the discounted value of the next coupon payment multiplied by the probability of no event
ii. If an event occurs before the next coupon payment at time , the recovery on the LIBOR floating rate note, which equals the discounted recovery of principal multiplied by the probability of the loss occurring at time s, summed across all times
iii. The value of a LIBOR floating rate note received at the next payment date , which equals the value of a LIBOR floating rate note A discounted to the present, multiplied by the probability of no event
iv. Less the expected loss after the next coupon payment, which is the discounted loss at time s, multiplied by the probability of the loss occurring at time s, summed across all times
v. Plus, the expected fixed payments after the next coupon payment
Hardle and Lopez Cabrera (2010[79]) evaluate the calibration of a real parametric CAT bond sponsored by the Mexican government to secure coverage against earthquake events and derive the price of a hypothetical modelled-index CAT bond for Mexican earthquakes. Annex Box 3.A.1 illustrates the main steps in the pricing of this hypothetical CAT bond.
Annex Box 3.A.1. Example of pricing approach for a hypothetical modelled-index CAT bond for Mexican earthquakes
Hardle and Lopez Cabrera (2010[79]) evaluate the pricing of a catastrophe (CAT) bond for earthquakes in Mexico with a modelled-index loss trigger mechanism. The authors apply the CAT bond pricing methodology of Burnecki and Kukla (2003[75]) and Baryshnikov, Mayo and Taylor (1998[74]) to the loss data obtained from the National Institute of Seismology in Mexico. The loss data contain information about 192 earthquakes with a magnitude above 6.5 on the Richter scale that have occurred in Mexico between 1900 and 2003. The historical losses were adjusted to the population growth, the inflation rate and the exchange rate (MXN [Mexican peso]/USD [United States dollar]) converted to USD in 1990. The losses are modelled by means of linear regression methods, using as explanatory variables the earthquake magnitude, the depth of the earthquake and the impact of the earthquake on Mexico City.
The authors then model earthquake arrival frequency through various methods and conclude that the process can be correctly modelled with a Homogeneous Poisson process. Next, Hardle and Lopez Cabrera (2010[79]) assume a zero-coupon CAT bond (ZCCB) that pays a principal amount P at time to maturity T, conditional on the threshold time τ > T. P is a predictable process, which implies that the payment at maturity is independent of the occurrence and timing of the threshold D. The principal P is fully lost in case of occurrence of a trigger event.
The non-arbitrage price of the ZCCB is associated with the threshold D, the earthquake flow process Ns with intensity rate λs and a loss distribution function F(x).
Following these pricing assumptions, the authors obtain the values of a ZCCB with principal P = 160 million at t = 0 with respect to the threshold level D and expiration time t ∈ (0, 3) years. The threshold D is defined as D ∈ [100, 135] million, which corresponds to the 0.7 and 0.8 quantiles of the three yearly accumulated losses (i.e. approximately three payoffs are anticipated to occur in 100 years). The authors further assume a continuously compounded discount interest rate to be constant and equal to the London Interbank Offered Rate (LIBOR) in May 2006 and an earthquake intensity rate =1.8504. When D = 135 million and T = 1 year, the ZCCB price is equal to 151.78 million.
Source: Hardle and Lopez Cabrera (2010[79]), “Calibrating CAT Bonds for Mexican Earthquakes”, The Journal of Risk and Insurance, https://doi.org/10.1111/j.1539-6975.2010.01355.x.
In a more recent paper, Nowak and Romaniuk (2016[80]) use the martingale method for pricing CAT bonds. More precisely, the authors price the CAT bonds by means of a generalised payoff structure, which assumes that the bondholder’s payoff depends on an underlying asset driven by a stochastic jump-diffusion process (i.e. a model that includes both stochastic volatility and jumps). Simultaneously, the risk-free spot interest rate also has a stochastic form and is described by the Cox, Ingersoll and Ross (1985[81]) model. Furthermore, Nowak and Romaniuk (2016[80]) assume the possibility of correlation between the Brownian part of the underlying asset and the components of the interest rate model.
Empirical approaches
Empirical approaches have emerged since the first CAT bond issues. They have the advantage of being more pragmatic than theoretical approaches and are better understood by investors. In addition, linear regression models are shown to perform relatively well in the out-of-sample forecast of CAT bond premiums (Trottier and Lai, 2017[82]). One of the main empirical approaches is proposed by Lane (2000[83]) and continuously improved. This approach only focuses on the risk load and establishes that the premium over the expected loss is a Cobb-Douglas-type function of probability and expected loss in the event of default. More precisely, Lane (2000) posits that the CAT bond premium can be derived according to the following formula:
Where PFL is the probability of first loss, CEL is the conditional expected loss, and EL is the expected loss. PFL and CEL take values between 0 and 1.
Instead, Lane and Mahul (2008[84]) accomplish a multivariate linear regression analysis in order to identify factors explaining the risk loads of CAT bonds in addition to the expected loss. They suggest a multiple linear relation between the premium, the underlying peril, the expected loss, the wider capital market cycle and the risk profile of the transaction. The simple linear model has the following form:
The authors then adjust the simple model to account for the cycle by using two approaches, namely add another coefficient to the regression or divide the actual spread by the index deflator. This results in the following two specifications:
Or equivalently:
Last, Lane and Mahul (2008[84]) also amend the simple model to allow for peril exposure as follows:
Where parameter is the coefficient associated with peril p, and is the expected loss from peril p.
Other approaches consider further factors that could drive the CAT bond premium, in addition to those outlined in the approaches above. Braun (2015[19]), for example, proposes a linear model that includes a dummy variable for exposure to peak territory, a dummy variable for being sponsored by Swiss Re and a dummy variable for being rated investment grade. The model contains no intercept to reflect the fact that the explanatory variables proxy CAT bond risk. All variables equalling zero would correspond to the risk-free rate. The multilinear model can be written as follows:
Some authors have departed from the linear model, opting instead for a log-linear specification. Papachristou (2011[32]), for instance, fits a Generalised Additive Model to market data, considering various explanatory variables for CAT bond prices. One variable is the expected loss that reflects, to some extent, the volatility of losses. Another variable is the type of peril and territories covered, mainly reflecting correlation with investor portfolios. In addition, the author includes the reinsurance cycle, reflecting loss experience, changes in risk perception over time and the availability of capital. Last, the type of trigger is also included, to reflect the amount of basis risk. The selected model has the following form:
Where is the risk load, is the expected loss, fs are smoothing functions, and are factor variables, and are independent and identically distributed random variables.
A log-linear approach is also outlined in Mevorach (2018[85]). The author develops a multivariate log-linear model that includes cyclical as well as CAT bond-specific variables that have been established in the literature and are highly statistically significant. The model has the following form:
Where EL is the expected loss, is the Lane Financial synthetic insurance-linked securities rate-on-line index, is a dummy variable for exposure to US perils, is a dummy variable for exposure to hurricanes, is a dummy variable for exposure to Japanese earthquakes, and are independent and identically distributed random variables.
Challenges of existing CAT bond modelling approaches
Examples of limitations of CAT bond pricing models
Modelling the valuation of CAT bonds is an area for continued research and development. Since the CAT bond market is incomplete (Galeotti, Gürtler and Winkelvos, 2012[86]), the deployment of an appropriate pricing model is of outmost importance. Current modelling approaches have various limitations. More specific and theory-based models are necessary, as most models used by practitioners are primarily descriptive rather than grounded in theory. In addition, the very existence of the risk premium is modelled on an ad hoc basis, with no economic explanation for its existence. In particular, the models do not consider the systematic nature of the risk associated with CAT bonds. Furthermore, since the global financial crisis, CAT bond spreads have become correlated with those of corporate bonds, suggesting the emergence of a significant systematic component in CAT bond prices. CAT bond investors are also more sensitive to catastrophe events that trigger a reassessment of risk (Carayannopoulos and Perez, 2014[87]).
Another thing to note is that the current models can be very sensitive to the parameters used, in the sense that a slight change in parameters can lead to significant losses. This issue could be addressed by establishing a maximum threshold of acceptable risk. An example are the floods that affected the United Kingdom in 2000. The floods caused severe damage, and insurance companies agreed to cover the losses. However, in return, the UK government took action to increase annual expenditure to enhance flood preparedness (Kelman, 2001[88]). Relatedly, the use of historical data for the estimation of future losses is rather problematic, as many variables affecting the size of these insurance losses have changed materially (Canabarro et al., 2000[89]). For instance, demographic changes and mitigating factors (e.g. improvements in construction standards and risk management strategies) mean that historical data may be of little relevance for forecasting future losses caused by catastrophe events.
Model calibration has significant practical importance. The appropriate calibration of a CAT bond model requires specifying a probability distribution for the underlying variables in such a manner that the model is able to reproduce the current market prices of a set of benchmark financial instruments. It is well known, however, that sometimes there can be multiple solutions, and sometimes there is no solution at all. This means that prices may not be consistent with any risk-neutral probability (i.e. an arbitrage exists) or that there exist several risk-neutral probabilities consistent with the benchmark prices due to market incompleteness. In an incomplete market, not all states of nature can be spanned, and as a result, parties are not able to move funds freely across time and space, nor to manage risk (Heaton and Lucas, 1996[90]). As the CAT bond market is incomplete (Galeotti, Gürtler and Winkelvos, 2012[86]), any calibration procedure involves making subjective choices.
Existing models may not be suitable for capturing natural hazards that are very common in Emerging Asian countries. For example, floods are one of the regular hazards in the region. In addition, countries also often struggle with slow-onset hazards, such as drought. However, the existing models are focused on either earthquakes or tropical cyclone events. In the case of drought, no probabilistic drought risk model for Emerging Asia currently exists (UN ESCAP, 2019[91]). Thus, an additional key step is adapting the models to the geographic area of interest.
Machine learning could improve the performance of CAT bond pricing models
A robust CAT bond pricing model should be based on both statistically and economically significant price determinants and avoid overfitting issues. Furthermore, a good forecasting framework must correctly specify the functional relationship between the dependent and explanatory variables and provide a suitable choice of the underlying conditions of the prediction (Gu, Kelly and Xiu, 2020[92]). The development of an appropriate forecasting model is therefore a complex undertaking, and existing theoretical and empirical models may not always provide a good solution in this regard. Artificial intelligence could considerably improve pricing models’ forecasting performance by allowing a rich set of possible model specifications compared to conventional methods (Gu, Kelly and Xiu, 2020[92]; Mullainathan and Spiess, 2017[93]). More recently, several studies have emerged, ranking the performance of various machine learning techniques in asset pricing models (Annex Table 3.A.1).
Annex Table 3.A.1. Examples of machine learning methods with the highest forecasting performance in asset pricing models
Source: Authors’ compilation, based on cited references.
Some authors evaluate the performance of these techniques in relation to CAT bond pricing. Makariou, Barrieu and Chen (2021[16]), for instance, apply the random forest method to predict spreads in the full spectrum of the primary non-life catastrophe bond market. They find that random forests have at least as good a prediction performance as the benchmark linear regression in the temporal context and superior prediction performance in the non-temporal one. The authors also conclude that the random forest approach performs better than the benchmark model when multiple predictors are excluded in accordance with the importance rankings or at random. This result suggests that the random forest method extracts information from existing predictors more effectively and captures interactions better without the need to specify them.
Likewise, Götze, Gürtler and Witowski (2020[15]) assess the forecasting performance of linear regression models versus machine learning techniques in the CAT bond market. The authors use linear regression with variable selection, penalisation methods, random forests and neural networks in order to forecast CAT bond premiums, concluding that random forests exhibit the highest forecasting performance. Additionally, random forests display a smaller variance in forecasting performance over time compared to the linear regression model. This result is important in a context where uncertainty is high and performance stability becomes highly relevant. On the other hand, Götze, Gürtler and Witowski (2020[15]) suggest that the performance of the neural network approach depends on the applied test specification and lags both the linear regression models and the random forests.
The literature on the performance of machine learning methods for asset pricing models in general is broader than that narrowly focused on the CAT bond market. These studies could nevertheless provide useful insights on machine learning methods that could also be applied to the pricing of CAT bonds. A number of authors use machine learning techniques to deal with the high-dimensionality challenge. Rapach, Strauss and Zhou (2013[94]) apply the lasso method to select a few predictors from a large set of candidates, while Stambaugh and Yuan (2016[95]) use covariance cluster analysis to identify two groups of related anomalies and then build factors based on stocks’ average within-cluster characteristics rank. Kelly, Pruitt and Su (2019[96]) perform dimensionality reduction of the characteristics space by extending the projected principal-component analysis to allow for time-varying factor loadings. In a comparative study of several machine learning methods for measuring asset risk premiums, Gu, Kelly and Xiu (2020[92]) conclude that boosted regression trees, random forests and neural networks are the best-performing methods.
Notes
← 1. Inflation-adjusted estimates. Most of this increase is attributable to economic growth, population growth and urbanisation, which lead to a higher concentration of assets in areas exposed to natural disasters.
← 2. The most important reinsurance renewal dates are 1 January (Europe), 1 April (Japan) and 1 July (United States).
← 3. For example, Swiss Re and Munich Re, the two largest reinsurers in the world, both exhibit financial strength ratings of AA- by Standard & Poor’s.
← 4. For example, a 2011 report by the National Association of Insurance Commissioners on classical insurance markets found that delayed indemnification can result in additional legal fees, interest charges and other expenses.
← 5. This bond relies on a parametric trigger, referencing data from the US Geological Survey.
← 6. The ILS markets recently witnessed the advent of CAT bonds lite, highlighting the fact that structuring and issuance costs are a major quantity. A cat bond lite structure relinquishes cost-intensive steps in the issuance process, such as a full prospectus and catastrophe model report. In addition, it is privately placed so that distribution costs are lower than for standard CAT bond structures. For further information, see: https://www.artemis.bm/news/cat-bond-lites-are-growing-in-number-and-importance/.
← 7. Prominent examples are the FloodSmart Re transactions by the US National Flood Insurance Program. For more information see: www.artemis.bm/deal-directory/.
← 8. Basis risk is also a disadvantage of modelled-loss triggers. These are quite rare in today’s market. Under a modelled-loss trigger, the payout of a CAT bond is determined by the output of a catastrophe risk model. Such models are usually provided by one of the large model vendors (AIR Worldwide, RMS and CoreLogic). Post event, the model is fed the key parameters of the disaster, which it then maps into an estimate of the corresponding losses.
← 9. Recent statistics on parametric insurance programmes in India show that the penetration of coverage is still low: 65% of the crop area remains uninsured (Agriculture Times, 2021[97]). One driver of this persisting protection gap is likely to be insufficient density of the measurement infrastructure.