Credit Suisse Insurance Linked Strategies

Optimising diversification in a maturing market

Hamlin Lovell

Reinsurance risk including CAT bonds has not been immune from the yield compression seen in most credit markets, but over the past eighteen months yields have started to pick up. To a degree this reflects a clutch of concentrated losses in California wildfires, US hurricanes and Japanese typhoons, which have been at least partly aggravated by climate change. Credit Suisse has observed increases averaging around 15% and up to 35% in premium rates in 2019, sometimes in conjunction with higher retentions also reducing risk for ILS investors. 

The average historical return on CAT bond indices has been annualised around 7% but this has been a tale of two halves, with higher returns during the first 15 years of the market. “When CAT bonds were first issued in the early 1990s, their yields included something of a novelty premium for being exotic. That inefficiency vanished in 2009-2011 as more capital flowed in. 

The big advantage of private reinsurance is that it covers a much wider spectrum of more granular risk, which allows for greater diversification and lower tail risk in portfolio construction.

Niklaus Hilti, head of Credit Suisse’s Insurance Linked Strategies division

Today, yields of 4-5% (before any event losses), are broadly comparable to high yield corporate debt, leveraged loans, asset backed securities, structured credit, or emerging market sovereign debt in USD,” says Niklaus Hilti, head of Credit Suisse’s Insurance Linked Strategies division, which runs circa US$7.3bn as of July 2019, and has been active since 2003.

Negative interest rates have a direct impact on those few CAT bonds denominated in JPY or EUR, because the collateral is in money market funds or short dated government securities, which do not have the floors that some structured credit products have. In practice most CAT bonds are now issued in USD. Paying a fixed spread on top of collateral earning floating rates means their interest rate sensitivity is minimal.

Niklaus Hilti, head of Credit Suisse’s Insurance Linked Strategies division

The risk premium on ILS is not completely detached from global financial markets: low and negative rates have increased the volume of capital searching for yield, and placed pressure on reinsurance premiums and CAT bond yields. “The risk premium on CAT bonds dropped roughly 50% between 2011 and 2018, and has picked up by 10-20% in 2019 so it is now perhaps 35% or 40% below 2011,” says Hilti. “Private reinsurance saw increases somewhat earlier. We do not expect returns to revert to historical levels now that the market is more mature and well understood however.”

Indeed, the nature of the repricing has also been different this time around. “Back in 2001, after the September 11 World Trade Center attacks, the entire market repriced. Then in 2005 after Hurricane Katrina, only the US market repriced. In 2017 to 2019, it has been much more granular with only certain regional markets within the US, such as Florida, repricing. Meanwhile Europe – which has not had a very severe winter storm in 20 years – did not reprice. The market is now becoming much more efficient and discriminating.”

(Incidentally, the Solvency II rules have no real impact on non-life ILS – which are covered in the rest of this article. They have much more impact on life insurance-linked securities and market risk assessment, according to Hilti. Credit Suisse has a strategy focused on longevity and mortality risk. “The potential for a liquidity mismatch between investor terms and assets can be managed with exit clauses and valuation clauses applying to potentially longer dated longevity exposures,” he says.)

$7.3bn

Credit Suisse’s Insurance Linked Strategies division runs circa US$7.3bn as of July 2019.

Correlation with financial markets 

The level of risk premium is to some degree linked to the other asset classes with which they compete for capital. The key benefit of the asset class is diversification as losses on ILS should be generally uncorrelated with conventional financial markets. Natural catastrophes follow various statistical distributions, but these are not generally linked with bear markets in financial assets. There are some scenarios that could see both causal or coincidental correlation however. “An earthquake in San Francisco could cause some financial losses for companies, including the tech sector, in and around Silicon Valley, though this would be a local rather than a global reaction,” points out Hilti. “And the 2005 Hurricane Katrina did take out some gas capacity and contribute to a spike in natural gas prices.” Beyond this a natural catastrophe might randomly coincide with a financial market event: the UK Great Storm of October 15-16 1987 was very close to the Black Monday equity market crash, on October 19. But in general, the distribution of losses should not be correlated with equities, bonds or credit.

Counterparty and credit risk

In 2008, some specific Lehman-wrapped CAT bonds saw losses but the structure of bonds has changed. Counterparty risk is substantially mitigated by collateralisation of principal on the bonds. It is also worth recalling that if multiple catastrophes did render a reinsurer insolvent, the liquidators might need to collect principal from CAT bondholders or buyers of private reinsurance. The investors would probably not be owed anything by the reinsurers, besides perhaps the last coupon or interest payment. Investors might miss a coupon payment under this scenario, but this risk is pretty low, as coupons are quarterly and quite small versus the notional principal exposure. Credit Suisse is diversified across over 100 counterparties in many regions at any one time.

Maximising diversification 

A relatively small part of the overall reinsurance market is securitised into catastrophe bonds: US$28.7bn as of June 2019, versus a private ILS market – collateralised retrocessions/reinsurance, ILWs (Industry Loss Warranties), and Sidecars – of circa US$72bn, and another US$362bn in traditional reinsurance, according to Credit Suisse (with direct insurance worth over US$5tn). “CAT bonds are mainly issued for the biggest catastrophes, whereas reinsurance covers the full spectrum,” says Hilti. The private reinsurance deals do not offer a significant illiquidity premium for the same level of standalone default probability. “Pricing has converged between the two markets as sponsors will play one off against the other to get the best deal,” Hilti continues. “The big advantage of private reinsurance is that it covers a much wider spectrum of more granular risk, which allows for greater diversification and lower tail risk in portfolio construction. 

“The CAT bond is a very tight and small market, with about 70% or 80% of risk in US wind and hurricane. There are only about 150 CAT bonds, which could possibly be used to construct a portfolio with an upside yield of circa 5% and a downside tail risk of circa 50-60%. By contrast, when structuring portfolios, we aim for a ratio of drawdown or tail risk to yield of two or three to one, generating better risk-adjusted returns. For instance, a portfolio offering upside of 17-20% per year might have modelled tail risk (for a once in 100-year event) of 40-50%. For 5% upside, the tail risk might be between 12% and 17%.” 

It is also possible to diversify by peril, region and trigger.

Man-made risks and cybersecurity 

Man-made risks, such as marine/energy, aviation/aerospace, crop, events/contingency insurance, industrial accident, political, operational, terror, cyber, engineering, workers’ compensation, and cyber risk are also potentially investible. “But most of these are not very attractively priced as they can easily be absorbed by the reinsurance industry,” observes Hilti. “Cyber risk is where we see most potential. This is one of the fastest growing categories of man-made risk. It is one of the largest risks in society and cannot be diversified as a global cyberattack could affect the whole world. This is a relatively young market and we want to quantify risks before deciding if it is in the best interests of investors.”

Diversification by peril

Portfolios can diversify by natural perils, which include seismic (earthquakes and seaquakes), wildfire, floods, wind (hurricane, typhoon, winter storm, tornado, hail, severe weather) and tsunamis. “A concentrated portfolio can have a more binary profile, but might outperform in a period when there are a larger number of severe events (such as Hurricanes Harvey, Irma and Maria). Roughly 70% of the yield on typical Credit Suisse ILS strategies comes from weather related risk with 30% of it from earthquake risk, which is a much lower probability and lower yield risk – earthquakes may occur with a frequency of every 70, 100, 150, 250, 500 or 1,000 years, whereas storms may occur every 5, 10 or 20 years,” says Hilti.

Diversification between and within regions

“Correlation is generally assumed to be zero between major regional risks (such as US hurricane, European storm, Japanese typhoon or California/Italy/Japan earthquake) but within major regions there clearly could be correlations. For instance, France, Monaco and Italy might be touched by the same earthquake; and the same storm could hit the UK and Europe. Storm Humberto in September 2019 – hitting both the Bahamas and the Portuguese Azores islands – caused losses for US and European insurance exposures. Indeed, though the probability of hurricanes has not increased by a statistically significant degree, we are seeing more intense and larger hurricanes that survive for longer outside the Tropics, so the probability that they hit Europe is higher. This is one source of potential correlation,” says Hilti. “But even within regional perils, such as US hurricane risk, the reinsurance market allows for micro-diversification within regions, such as the North East, Florida, the Gulf, and nationwide.”

Diversification by trigger 

One axis of diversification balances out severity and frequency risk; frequency risk helps with diversification so that losses are not too large in an individual year. This is also important to allow for faster recovery of losses. “Frequency risks – second and subsequent event exposures – provide extra degrees of freedom of diversification. Instead of all risk being exposed to a first event structure, it could be spread by requiring at least two hurricanes,” he says. “This adds diversification potential even in the same risk region. There can be a range of junior to senior risks, and first to subsequent risks.”

Different types of loss triggers – indemnity triggers, industry loss triggers, second generation parametric triggers and modelled loss – have advantages and disadvantages including in terms of the time taken to determine losses, the quality of loss reporting, uncertainty over losses, and basis risk between the event and the insured loss. “For instance, an indemnity is a relatively complex trigger mechanism,” he continues. “An industry loss trigger can be faster, but is only as good as the reporting agent or index provider, and there can be uncertainty partly as some insurers may hedge part of their risk. Parametric triggers can be the fastest and cleanest way to establish insured losses based on an index value measuring physical variables. Basis risk can arise if a parametric trigger is based on earthquakes, but losses are caused by a seaquake or tsunami.” Basis risk could work both ways for investors: ILS that does not pay out to an insurer could of course be advantageous for the investor.

Deal sourcing 

It is easier to obtain most of these more granular risks and additional dimensions of diversification through private reinsurance than through CAT bonds. A typical split for Credit Suisse ILS is 50% CAT bonds, and the other half in private reinsurance with 20% from captive insurers, and 30% from third party insurers. A diverse sourcing network helps to access a wider variety of risks.

Credit Suisse sources private reinsurance and traditional reinsurance deal flow partly through four captive insurers: Bermuda based Bernina Re; Guernsey based reinsurers, Kelvin Re and Humboldt Re, and the Arcus Syndicate 1856 at Lloyds of London, which can also access direct insurance. They are independent companies but source transactions exclusively for the Credit Suisse fund platform, which also sources from third parties. 

Credit Suisse finds all three of these insurance jurisdictions are good places to do business. “Guernsey has a big fund platform and insurance market,” Hilti explains. “We like the expertise, proximity of the infrastructure and insurance platforms. Bermuda offers very good market access, particularly for US reinsurance risk. Guernsey and Bermuda both offer outsourced facilities to save us on costly administration. London is a very old and well-established market. Guernsey and Bermuda both have well established laws and regulations that understand how insurance linked securities work, which is different from insurance. London has also just introduced a legal framework in 2018.”

Consensus and dissensus in modelling risk 

Assessing catastrophe risk is subject to differences of opinion between risk modelling firms, and sometimes Credit Suisse also have a view that differs from more than one modelling firm, but this should not be exaggerated. “The major risks have a high degree of consensus, and generally minor risks are subject to dissensus,” says Hilti. “For consensus risks, benchmark or lead models for each region are updated annually based on new scientific findings. We carry out model verification to find the most suitable ones. A normal or lognormal distribution is clearly not appropriate for fitting to natural catastrophes. An extreme value distribution, such as Pareto or generalised Pareto, is a closer fit to historical data, but we do not actually do that beyond model verification. Instead we use stochastic models that start with historical events and then make small disturbances to their location thousands of times to arrive at a 100,000-year distribution that includes more of the tail risk. This Monte Carlo simulation process is more realistic, otherwise there are not enough storms historically to reach a conclusion.” 

“Australian earthquakes are an example of a risk with less consensus. There have been very few historical earthquakes and known records are not as long as in Europe. But these risks are in the tails anyway and play a much less important role for the economics of the product. Fundamentally, in the major regions of the US, Europe and Japan, there is a high level of agreement between the models.”

Is climate change risk being fairly priced?

“In general climate change warms the sea, which is where hurricanes get their energy from,” he continues. “It is undeniable that climate change is increasing the probability of water-related catastrophes, such as droughts, and pluvial (rain) floods (incidentally, floods are often not covered by insurance eg in Europe). The vast majority of scientists on the International Climate Change Panel agree on these things. Higher temperatures also increase the risk of heatwaves, droughts and wildfires. Heating up the atmosphere creates more water going up, which must come down, so there are more strong rainfalls, and hailstorms. The impact on US hurricanes is mixed.”

He continues that, “After a recurring pattern of these events, they are starting to get priced in and risk premiums – which are reset annually – are adjusting to reflect the increased risk. US wind has also been repricing regularly as the risk assessment has gone up every year for the past 15 years. There has been a very significant repricing. The probability doubled from 1 to 2% after 2005, and is now 3%, but the yield has gone from 5% to 7%. In contrast, European wind risk has been repriced downwards partly due to the absence of an event.”

Hilti broadly thinks that ILS markets are adjusting to account for the increased probabilities of some events, albeit with some time lag. “In contrast, infrastructure – which is also vulnerable to some natural catastrophes – can have returns locked in for up to 30 years and is therefore more exposed because it cannot adjust pricing.”

Are hurricane and earthquake frequencies increasing?

Notwithstanding increased risks of some weather events, Credit Suisse argues that the risks of the biggest two natural catastrophe categories – US hurricanes, and earthquakes globally, have not increased. Investors should not infer from record breaking financial losses that the physical frequency and severity of certain catastrophes has necessarily increased. Credit Suisse suggests that the frequency of hurricanes as a whole has not increased since 1851, according to the United States’ NOAA/National Weather Service (though the frequency of major hurricanes has slightly increased). Similarly, the frequency of worldwide earthquakes of magnitude 7 or higher has not increased, according to the United States Geological Service. “In both cases, increasing insured losses are mainly caused by the increase in building density, population density, insurance density, inflation, and wealth in the key areas. Building quality has improved, with better building codes, but there are more buildings in exposed regions,” says Hilti. Credit Suisse argues that if a simulation exercise maps historical hurricanes onto current levels of building, population and insurance density and values, recent losses are not unprecedented.

Weather forecasting and loss forecasting

The firm also has some confidence in forecasting event probabilities for individual years. Hilti, who is a meteorologist, says “forecasting weather is now better than guesswork. We wrote articles in 2003-2004 with Mark Saunders and Benjamin Lloyd-Hughes of University College London, on seasonal forecasts of hurricanes, typhoons and tropical storms, which forecast if a particular season would be active or not. Over long periods of time, it is possible to add value by lowering exposure if a very active season is forecast, and vice versa.” Credit Suisse has access to both licensed vendor risk models and proprietary tools, which may use aviation and satellite data.

Once an event does occur, risk assessment and quantification of losses is getting faster. Hilti recalls how “back in 1992, after Hurricane Andrew in the US, the area was closed for some time and insurance people had to wait four weeks to get access. Now, you would immediately get satellite and drone data to assess the real time impact. We saw that with the recent Typhoon Hagibis in Tokyo, where high resolution data showed building elevations, and measured floods.”

Both forecasting and faster risk assessment can help to inform a limited degree of active trading.

Active trading 

The strategy is long biased and mainly buy and hold over a one-year time frame, but there is limited scope for active trading of risk. The fact that most private insurance contracts last a maximum of 12 months allows for portfolio rebalancing around renewal dates. The biggest renewal cycle takes place in January of each year, Japanese contracts are renewed in April and US wind plus most Australia and New Zealand contracts are renewed in July. “Over the past few years, Credit Suisse has reduced European weightings as margins have contracted due to the absence of a storm while the Asia Pacific weighting has increased to capture larger premiums,” says Hilti.

In between renewal dates, there can be opportunities for some active trading, mainly by hedging risks during and after events. “We would not buy a distressed CAT bond because there is too much uncertainty over potential losses. But we might selectively go short of a particular catastrophe risk at various stages in the evolution of a natural catastrophe, through buying various types of hedges on an event. We are minded to hedge risk because we have observed a tendency to understate risks, perhaps because extrapolation is the easiest way to forecast. For instance, Orlando in Florida has seen a 40% population increase since 2000, so it was not accurate to use the last hurricane to estimate losses from Hurricane Dorian in 2019. We bought industry loss hedges to reduce the impact of Hurricane Harvey in 2017. Though these hedges are not huge trades, due to the lack of liquidity in the market, they can play some active role in hedging out risk.” Credit Suisse has not taken an outright short position in a CAT bond or reinsurance contract.

Absent any events, the main source of volatility is seasonal fluctuations. “For private reinsurance, most of the premium accrues during the relevant event season, such as the hurricane season (valuation is carried out independently, by external actuarial consulting firms). CAT bonds, which have a secondary market price, in practice display similar seasonal pricing patterns: they accrue the premium in a linear way, but mark to market on the pricing of the bond compensates,” says Hilti. Seasonal trading is seldom a realistic proposition, however. “In theory, you could sell a CAT bond before the hurricane season, but in practice you would not be able to buy it back after the season, and so would be under-invested. On one occasion we acquired a CAT bond in April that was bid up to 106 by July, when it should have been trading down to 95 or 86 ahead of the hurricane season. We seized the chance to make a swift profit, but such opportunities are rare for a strategy that is seeking to earn risk premiums.”

Black Swan natural catastrophe scenarios 

Credit Suisse’s 100,000-year simulations can identify potential for a hurricane to hit New York City and also the risk of a catastrophe impacting multiple major regions. “Potential sources of super extreme tail risk, or Black Swan catastrophes, are not considered by insurers who measure risk on a 200-year view under Solvency II,” says Hilti. “But they might occur within a 100,000-year period based on our modelling. These could also impact more than one continent. For instance, there is a big fault line running through Oregon, Vancouver and Alaska, which could create seaquakes of magnitude 8 or 9, resulting in a tsunami covering the US, Canada, Alaska and even Japan. The Canary Islands have a volcano which, if it erupts, could create a mega tsunami over the entire East Coast of the US. And a large, systemic meteorite hitting land or water could also reverberate globally, with much greater impacts than the meteorite that hit Siberia three or four years ago destroying many buildings,” he declares.

Another perspective on extreme risks comes from the oldest works of human literature. The story of Noah’s Ark in the Biblical Old Testament certainly sounds more severe than anything in living memory. If these stories are interpreted literally, and presumed to have been passed down the generations over the estimated 300,000 years of human existence on the planet, then there have been super-extreme catastrophes. Under these scenarios, basic survival could become more important than financial losses. 

Perhaps insurers could work with anthropologists and linguists to build bridges with indigenous peoples, such as Australian Aborigines and New Zealand Maoris, to gain deeper insights into the events of the past.

ESG 

“Broadly speaking, the ILS market is ESG friendly, because it supports a chain of risk transfer, which sends capital to places where people have suffered from big catastrophes and losses. As such it is supported by governments, and intergovernmental institutions,” says Hilti. 

Credit Suisse has a firmwide ESG policy which applies to the ILS business. “This does not carve out insuring terrorist related risks. While we are not exposed to any standalone terrorist risks, it is not in practice possible to completely exclude terror. Similarly, offshore energy is not perceived as a sustainable industry, but cannot realistically be carved out of a catastrophe bond that insures certain regions which include offshore oil and gas production. Overall, these are tiny or non-existent risks,” says Hilti.

The aspiration to use CAT bonds for “Impact Investing” is seen in some issues sponsored by the World Bank and development banks. For instance, in April 2018 it issued US$1.38bn CAT bonds providing earthquake cover in Chile, Colombia, Mexico, and Peru. In October 2019, the World Bank started helping Jamaica to pay premiums on a small CAT bond. In total the World Bank has sponsored around US$4bn of risk transfer bonds, so the market is currently a small part of ILS. Indeed, there is generally very little ILS in emerging markets. “CAT bonds are mainly issued for risks in the US, Europe and Japan, with very little in emerging markets where risk assessment is much more uncertain, and where there is less insurance density as well, so the insurance industry has not invested enough time and resources in developing models,” says Hilti. “In theory, microfinance insurance could be securitised but in practice, we have not seen any such bonds and this is probably because the volume of risk is currently too small to require offloading onto capital markets,” he adds. It remains to be seen if emerging and frontier markets develop a large and liquid enough ILS market to allow for capital to be deployed in a diversified way in these risks.

No product is being offered and this information does not constitute an offer to sell, or the solicitation of an offer to buy, any security, product or service of Credit Suisse Insurance Linked Strategies or its affiliates. It is not investment research or a research recommendation for regulatory purposes as it does not constitute substantive research or analysis. This material is provided for informational and illustrative purposes. It does not constitute an invitation or offer to the public to subscribe for or purchase any product or service. The information contained in this document has been provided as a general commentary only and does not constitute any form of regulated financial advice, legal, tax or other regulated financial service. Credit Suisse Insurance Linked Strategies accepts no liability for any loss arising from the use of this material.