Whether 2001 will prove to be the year of the catastrophee bond.
Expectations are that 2001 will be a breakout year for the catastrophe bond market, with issuance projected to double to $2bn. Two events have formed the catalyst for this market growth: high 1999 industry losses coupled with contraction in high quality retrocession capital. Reinsurance companies are expected to lead the issuance wave, while primary insurers are expected to follow suit due to the recent increase in reinsurance rates. On the investor side, catastrophe bonds are continuing to broaden their sponsorship due to their strong historical performance and improvement in liquidity. Moreover, investor interest in non-correlated assets has grown dramatically over the past year amid the weak performance of other financial markets.
We expect catastrophe bond issuance to rise to $2bn in 2001, up from an annual average of $1bn over the 1997-2000 period. The reason for the heavier volume of issuance is clear: changing industry economics have made so-called CAT bonds a more cost-efficient form of risk transfer. CAT bond issuance is largely dependent on the price of alternate forms of loss coverage for both primary insurance companies and reinsurance companies. Primary insurance companies increase their catastrophe bond issuance when reinsurance rates rise. Similarly, reinsurance companies issue CAT bonds when high quality retrocession – basically reinsurance for reinsurance companies – becomes more scarce and/or expensive. With reinsurance and retrocession rates currently on the rise, primary insurance and reinsurance companies are focusing attention on the capital markets for risk transfer .
Over the past few years, CAT bond issuance has been suppressed by the low prevailing rates in the reinsurance markets. The reinsurance market is a cyclical business: rates tend to rise following significant industry losses, then decline during periods of low losses. These fluctuations can be observed in the Paragon Catastrophe Price Index, which is an index of US reinsurance rates across various catastrophic risk classes. Reinsurance rates soared during the mid-1990s as insurance risk was repriced following the high industry losses from Hurricane Andrew and the Northridge earthquake in. However, rates subsequently dropped by 35% over the 1994 to 1998 period as the reinsurance industry enjoyed low insured losses and strong returns to their investment portfolios. While January 2001 renewal rates are not yet publicly known, it is widely believed that prices have risen by 10-15% over the past year.
This market hardening is also observable in indicative prices of industry loss warranties (ILWs), which are an alternate form of hedging against high industry losses (figure 2). The payoff structure of a typical ILW is fairly straightforward: a premium is initially paid for the ILW, which subsequently pays out if industry losses cross a specified threshold in a given year. For example, an ILW against $20bn US industry losses demanded a 10.0% premium in November 1998 (lower left corner of table). The January 2001 market indications show a marked increase in ILW pricing over the past year. The premium for a $20bn ILW, for example, has risen to 15.5%. These indications may overstate the increase in reinsurance rates, as the ILW market is thinly traded. Nevertheless, we view this evidence as being consistent with a general hardening of the reinsurance market.
The 1999 calendar year was the second most expensive year on record for the insurance industry, with over $28bn in insured catastrophe losses worldwide. Since most of these losses occurred in late December 1999, they have continued to drive up reinsurance rates during the 2001 renewal period. Moreover, many reinsurance companies locked in rates for two years in 1999 due to Y2K fears, and therefore have only recently rolled off. Losses in 1999 were just shy of the 1992 record of $29bn, when Hurricane Andrew alone caused $19bn in insured losses. Moreover, 1999 industry losses dwarfed the 1994 total of $21bn, which was boosted by $14bn of damages caused by the Northridge earthquake. Unlike 1992 and 1994, both marked by a single large catastrophic event, 1999 was marked by an aggregation of several smaller catastrophes. More recently, insurance industry losses in 2000 were relatively benign at just over $7bn, while damage estimates for the recent Indian earthquake are as high as $5bn, although insured losses will likely be much lower.
Only a quarter of worldwide losses occurred in the US in 1999, compared with a 15-year average of over 50%. Of the five natural disasters with insured losses over $2bn, only one – Hurricane Floyd – occurred in the US (figure 2). Two of the worst natural disasters in 1999 were European winter storms Lothar and Martin, which occurred back to back in late December. These non-tropical cyclones caused considerable destruction in France, Germany and Switzerland, and cost insurers a combined $6.7bn. Since reinsurance is a worldwide market, their impact continues to be felt by reinsurance companies around the globe.
Recent high industry losses have had their greatest impact on the retrocession market. As a result, the pricing of reinsurance and, ultimately, primary insurance is partially driven by the cost of retrocession. High losses over the past two years – both catastrophe and non-catastrophe – have considerably weakened the retrocession market. This has created a snowball effect: reinsurance rates have risen, primary rates have subsequently risen, and both reinsurance companies and primaries are looking to the capital markets as a form of risk transfer.
Retrocession rates were historically low during the mid- to late-1990s, a period when retrocessionaires expanded their portfolios through aggressive pricing. This aggressive pricing strategy has backfired over the past two years. Cumulative losses due to numerous smaller catastrophes over the past two years have crippled many of the weaker retrocessionaires. Non-catastrophic losses also impacted the retrocession market, which holds risks across various insurance markets. The worst non-catastrophic losses were related to Unicover, which pooled US workers' compensation exposure and sold health/medical claims to reinsurers and retrocessionaires. Inadequate pricing of Unicover-related risks resulted in $2bn of losses and the downfall of Reliance Group Holdings Inc.
The recent shakeout in the retrocession market has resulted in a significant reduction in capital and a reassessment of the risk-reward trade-off in the retrocession market. As a result, retrocession prices have increased. Reinsurance rates have subsequently risen due to the increasing cost of retrocession. Moreover, reinsurers have increasingly restricted their retrocession coverage to the better capitalised retrocessionaires, which typically charge higher rates. Due to the rising reinsurance and retrocession rates, catastrophe bonds have become economically attractive to both primary insurance and reinsurance companies.
Cash flows to a catastrophe bond depend on a trigger event, which determines when there is a loss of principal to the investor. There have been four basic types of triggers employed to date, all of which were employed over the past year:
l indemnity – indemnified notes have triggers based on the actual incurred losses on a pre-specified group of policies. While indemnity transactions provide the best insurance coverage for the ceding company, they require additional investor due diligence as they depend on the underwriting and claims settlement practices of the ceding company. Investors are also exposed to potential extension during the period when damage claims are settled following a catastrophe, known as the development period, which could potentially be up to two years. This class of catastrophe bonds includes the Residential Re series of transactions, which have been issued annually since 1997;
l model-based – these transactions have cash flows that depend on loss output from a third-party model analysis of an event's input on a notional portfolio. This portfolio is built to mirror losses to the ceding company's portfolio given a peril's parameters. Since these transactions are model-based, the development period is substantially shorter than in an indemnity transaction. The Mediterranean Re transaction issued in November 2000 is a model-based transaction that securitises some of AGF's exposure to French windstorms and Monaco earthquakes;
l parametric – cash flows for parametric notes depend on physical parameters of the event causing insurance losses, such as hurricane windspeed or earthquake magnitude. These transactions can be tailored using a formula-based index to better match the cedant's insured portfolio. Parametric bonds include the PRIME transactions sponsored by Munich Re in December 2000, whose cash flows were based on parameters such as central pressure and location of landfall of a Florida/New York hurricane, moment magnitude, depth, and location of epicentre of a California earthquake, and an index of windspeeds across Europe; and
l index – indexed notes have triggers based on a specified insurance loss index. The most commonly used indices are produced by Property Claims Services (PCS), who tracks US insurance losses by region and peril. Indexed securities are simpler structures for investors as they do not expose investors to idiosyncrasies such as portfolio concentrations. However, the cedent becomes exposed to basis risk between its actual portfolio losses and the loss index used in the transaction. The Seismic Ltd transaction issued in March 2000, which was based on PCS estimates of insured losses related to a California earthquake, is an example of a recent index-linked transaction.
Over the past year, issuance has shifted away from indemnity triggers and toward parametric and index triggers. Only 44% of last year's CAT bond issuance had indemnity triggers, compared with roughly 75% of 1997-1999 issuance. In contrast, parametric or index triggers were employed in 41% of 2000 transactions, compared with only 14% of 1997-1999 transactions. The low issuance of indemnity transactions is largely due to data constraints. Investors demand high quality underwriting data when purchasing an indemnity-based CAT bond. However, reinsurance companies have limited access to underwriting data from primary insurance companies, especially for non-US primaries. As a result, reinsurance companies are often forced to issue parametric or index-based transactions, exposing them to basis risk. We expect indemnity issuance to rise as high quality indemnity data becomes more widely available.
Greater focus on multi-year
CAT bond issuance has increasingly shifted toward multi-year issuance over the past year (figure 3). CAT bond issuance initially was concentrated in relatively short maturity bonds. In 1997, for example, all of the issuance was under two years of maturity except the ten-year Parametric Re transaction. This has changed dramatically over the past four years. In 2000, more than two-thirds of the issuance was three years or longer in maturity. Both investors and issuers clearly benefit from multi-year issuance. While investors gain the ability to amortise their due diligence over a longer period of time, issuers are similarly able to amortise their up-front underwriting costs. In addition, secondary trading has improved with a greater supply of outstanding issuance in the market.
On the investor side, sponsorship for catastrophe bonds has broadened considerably over the past few years. While some of the earliest CAT bond transactions were sponsored by a handful of investors, the more recent transactions have typically seen participation from dozens of investors. Distribution now spans across all investor classes, with significant participation from money managers, banks, insurance companies, and reinsurers. Hedge fund participation, on the other hand, has declined markedly with the demise of several significant players in this sector. Across all investor groups, portfolio managers increasingly recognise the value of the non-correlated aspect of CAT bonds, particularly after the credit debacles of 2000 in the corporate and high yield sectors.
The CAT bond market has begun to show the hallmarks of a maturing market. Tiering is becoming more important: investors now differentiate between various types of transactions. Spreads are typically tighter for single peril transactions versus multiple perils. Investors demand stronger underwriting and perform more data analysis on indemnity deals than other types of transactions. Investor confidence in the modelling of the underlying risks has increased, as CAT bonds have never suffered a loss despite the occurrences of Hurricane Floyd and the European windstorms Lothar and Martin. In the secondary market, trading volume has risen while bid/ask spreads have tightened.
In the current economic environment, the most compelling attribute of CAT bonds is zero correlation: their excess returns are not correlated with those of other financial assets. This feature makes CAT bonds unique among financial assets. During periods of financial uncertainty, other risky financial assets tend to underperform their riskless benchmarks as investors rotate into treasuries (figure 4). Over the past decade, the financial markets have been rocked by various geopolitical events (e.g. the Gulf War), credit/liquidity events (e.g. Russian default), or pure credit events in the corporate bond/high yield market. Returns on typical risky financial assets – including corporate bonds, high yield debt, emerging markets debt and equity – all produced significantly negative returns during these periods. Indeed, the positive correlation between asset returns tends to increase during credit downturns, further diminishing the benefits of portfolio diversification when it is most needed.
On the other hand, the risk of a natural disaster is perfectly uncorrelated with events that typically affect the capital markets, such as wars, economic slowdowns, rising corporate/consumer defaults, weak consumer confidence, Federal policy changes, etc. From a financial economics viewpoint, this expands the efficient frontier for investors, allowing them to earn more average return per unit risk. As a result, CAT bonds have performed well during past financial crises. For example, Residential Re bonds did not trade below 99.50 during the fall of 1998 despite potential of large supply hitting the market from liquidating hedge funds. Similarly, demand for CAT bonds remained strong during late 2000, when rising credit concerns curtailed demand for corporate and high yield debt. Indeed, the $300m CAT bond issuance placed by Munich Re in December 2000 was one of the largest fixed income issues in its rating class during the last few months of the year. CAT bonds are a particularly attractive investment in the current weak economic environment, which is rife with recessionary concerns. Moreover, since CAT bonds typically offer LIBOR spreads that are comparable to similarly-rated corporate bonds, investors do not need to give up spread to gain the benefits of zero correlation.
The expansion of securitisation to new perils has increased the benefits of non-correlation, as investors can benefit from portfolio diversification within the CAT bond sector. Issuers initially focused on securitising the most common risks: hurricanes on the eastern seaboard of the US, California earthquakes and Tokyo earthquakes. The list of securitised perils has lengthened considerably over the past two years, and now includes over a dozen independent perils. The newer securitised perils include exposure to earthquakes in Monaco and well as along the New Madrid fault, the most significant US fault zone east of the Rockies. In addition, the combined impact of Lothar and Martin has increased securitisation of European windstorms, notably in France and Germany.
A diversified portfolio of single-peril CAT bonds with independent risks can greatly improve the CAT bond return profile. One of the key concerns among new CAT bond investors is the so-called risk of ruin: the small probability (typically between 0.5% to 1% of the time) that an investor can lose their entire principal due to a single catastrophe. Since the occurrence of more than one independent catastrophic event is extremely unlikely, an investor can minimise the likelihood of a total principal loss by holding multiple independent single peril bonds.
From a statistical point of view, diversification within the CAT bond sector effectively smoothes out the fat lower tail of the CAT return distribution, reducing the portfolio's variance, skewness, and kurtosis. As an example, we consider the following portfolio of five recently-issued single-peril CAT bonds whose returns depend on five independent earthquake and hurricane/typhoon risks (figure 5).
Investors who hold only one CAT bond will lose some principal (attachment) between 0.94% and 1.47% of the time. They will lose their entire principal (exhaustion) between 0.32% and 1.08% of the time. An equally-weighted portfolio of these five CAT bonds will lose some amount of principal more frequently – roughly 5.6% of the time. However, the likelihood of losing more than 20% of the portfolio's principal is less than 0.4%, while the probability of losing more than 40% of the portfolio's principal is infinitesimal. For many investors, the relatively low likelihood of realising a large principal writedown makes an investment in a portfolio of CAT bonds considerably more appealing.
After several years of relatively light issuance, the CAT bond market is poised for rapid growth in 2001. The changing economics of the reinsurance industry – particularly the decline of the retrocession market – has made CAT bond issuance attractive to both reinsurance and primary insurance companies. Moreover, the investor base for CAT bonds is expanding due to strong historical performance, improving liquidity, and growing interest in non-correlated assets. With 2001 issuance projected at $2bn, we believe that the CAT bond market should demand increased investor focus over the next year.