Saving the world, one bond at a time

In March 2014, the government of Guinea reported an outbreak of Ebola to the World Health Organization (WHO). From 49 cases, there had already been 29 deaths. The disease quickly began to ravage west Africa, but it was not until October of that year, fully six months after the first case was reported, that sizeable amounts of money began to be put to work to contain the outbreak. In just the preceding three months, the number of cases had risen 10-fold.

When it finally came, the financial assistance was big – but arrived only after the disease had devastated local populations and economies. The scale of the outbreak was such that some $7 billion ended up being committed by donor countries. More than 11,000 people died. Had even a much smaller amount of financial assistance been available much sooner, the total required to get control of the situation, and the human cost in lives lost, might have been far smaller.

The difficulty of quickly mobilising assistance was a wake-up call for international institutions, including the World Bank, which estimates the total economic impact on the worst affected countries – Guinea, Liberia and Sierra Leone – to have been in the order of $2.8 billion. The World Bank was determined that developing countries should never again be left so unable to cope with an outbreak of this scale.

Independent expert panels convened since the Ebola outbreak have stressed the need for a global push to improve pandemic response. Part of that involves designing systems that actively encourage stricken countries to report crises as early as possible rather than delay notification until a time when it becomes harder and more costly to respond. One recommendation from the Ebola Interim Assessment Panel commissioned by the WHO stated that such incentives “might include innovative financing mechanisms such as insurance triggered to mitigate adverse economic effects”.

It was against this backdrop that the World Bank developed the Pandemic Emergency Financing Facility (PEF), a $500 million scheme that it announced at the G7 Finance Ministers meeting in Sendai, Japan, in May 2016. The most important coordinated measure previously available was the WHO's Contingency Fund for Emergencies (CFE), which is deployed to support the immediate response to a crisis.

The PEF, by contrast, is designed to funnel financial assistance to an outbreak of disease as soon as certain severity triggers are reached. It is intended to help prevent a pandemic from taking hold. Munich Re, one of the structuring agents of the facility, says it is designed to “fill the pandemic-response funding gap before large-scale disaster and humanitarian relief funding is mobilised”.

The World Bank reckons that if the PEF had existed in 2014, about $100 million could have been mobilised from it as early as July of that year – just a few months after the first reports of the Ebola outbreak. World Bank president Jim Yong Kim says that the facility “has the potential to save millions of lives and entire economies from one of the greatest systemic threats we face”.

The PEF has what the World Bank calls an "insurance window" and a "cash window", with premiums funded by Japan and Germany as donor countries and the cash part funded by Germany. The final step in setting up the facility was to transfer the risk to the public market. On June 28 this year, the World Bank priced $320 million of pandemic bonds to do just this, alongside derivatives transactions that take the total risk transfer to $425 million.

The deal borrowed from traditional catastrophe risk transfer solutions, but also set many precedents. It is the first time that developing market pandemic risk has been transferred. More importantly, the design of its pay-out structure is unique. It may not remain so: those who worked on the trade reckon it might well point the way for the future of the entire catastrophe sector. 

In insurance circles, pandemics are often described as among the most certain of uninsured risks. The World Bank estimates that the annual cost of moderately severe to severe pandemics is about $570 million, or 0.7% of global income.

This doesn’t sound like much, but annual averages underestimate the damage when a big pandemic hits. A pandemic on the scale of the Spanish Flu outbreak of 1918 to 1920, which is estimated to have killed as many as 100 million people, might have cost as much as 5% of global GDP, according to the World Bank. The same scenario could see the life insurance industry face losses of up to $50 billion, according to AIR Worldwide, a catastrophe-risk modelling firm.

A 2008 paper by David Rains, a managing director at Guy Carpenter, an insurance and reinsurance solutions provider, outlined the difficulties inherent in managing the risks of low frequency, high severity events such as pandemics. And although the market has moved on since then, his summary of the challenges still largely applies.

"The probability of an event occurring in any particular year is low," he wrote. "Even if an outbreak does occur, the process for estimating losses and determining reserves is unclear. Capital approaches do not consider probabilistic tail scenario risks. Quite simply, managing pandemic risk is an effort mired in doubt, though the potential for a devastating, multibillion-dollar, worldwide outbreak is real."

In such a scenario, insurers and reinsurers could face collapse if they carried meaningful exposure. The solution, argued Rains, was to "push the risk out of the (re)insurance market", via the capital markets. Extreme mortality bonds are one way that this has been done, typically using structures that are linked to a mortality index. Such an index has often been linked to existing data, such as that collated by the Centers for Disease Control and Prevention (CDC) in the US. When mortality goes past a certain point, the deal pays out to the issuer and bondholders lose principal.

Although it marks the first time that pandemic risk has been insured in developing markets, the PEF – and the way in which the risk is being transferred – draws on concepts from the catastrophe risk and extreme mortality market, as well as from previous work by the World Bank. The World Bank supports a Caribbean Catastrophe Risk Insurance Facility (CCRIF) to give coverage against natural disasters in the region. The Pacific Catastrophe Risk Assessment and Financing Initiative (PCRAFI) does the same for five Pacific island countries, while the Turkish Catastrophe Insurance Pool, a homeowner earthquake insurance scheme, was originally backed by the World Bank when it was established in 1999.

A bond issue in 2014 to back the Caribbean facility was the World Bank’s first cat bond, with the proceeds used to reinsure the scheme. That deal, a $30 million private placement, was innovative for skipping the MultiCat programme that the World Bank had set up in 2009 – a platform that issuers could use to sell bonds through a special purpose vehicle – in favour of direct issuance by the International Bank for Reconstruction and Development (IBRD), another World Bank institution, via its new capital-at-risk notes programme. The result was a simpler and cheaper process for the ultimate beneficiary of the proceeds.

There are precedents elsewhere too – in 2015 Axa Global Life sold an extreme mortality bond to transfer pandemic risk in France, Japan and the US, for example. But this market is still small. Issuance of extreme mortality catastrophe bonds has totalled about $3.5 billion since 2003. The broader insurance-linked securities (ILS) market is much bigger, however, and growing apace. The PEF deal takes public issuance in 2017 beyond the $8.2 billion annual record set in 2014, with practitioners expecting the year to easily surpass $10 billion.

×

But the new World Bank bond is novel in two ways. First, it transfers developing markets pandemic risk. And second, it is designed to pay out to the affected countries while an outbreak is taking place, with the goal of mitigating the eventual impact. That differs from structures that pay out after an event with the intention of replenishing insurers' capital. It also adds hugely to the complexity of designing the facility and marketing the bonds, as investors must be comfortable with the parametric triggers on which the pay-outs depend.

The PEF has two ‘windows’ – one with insurance, which is available now for an initial period of three years, and one with €50 million of cash, which will be available from 2018. It doesn’t cover every type of illness, but it covers six viruses that are considered the most likely to cause pandemics: influenza, coronaviruses (including the Middle East and Severe Acute respiratory syndromes, MERS and SARS), filoviruses (including Ebola and Marburg), Lassa haemorrhagic fever, Rift Valley fever and Crimean Congo haemorrhagic fever.

The assistance is specifically targeted at countries that are members of the International Development Association (IDA), a World Bank institution that assists poor countries, covering 1.6 billion people.

Most importantly, it is designed to shift financial assistance for emergencies away from what the World Bank’s Kim describes as “the cycle of panic and neglect” that he says has characterised the previous approach to pandemics. In a separate briefing on the PEF, the institution also notes that “time and again, the world continues to follow the same pattern: money isn’t brought to the table until a major outbreak hits an explosive point”.

The cash component of the facility, which has initial funding of €50 million, kicks in primarily to provide assistance for diseases that fall outside the six areas covered by the insurance scheme, or for situations that otherwise fail to meet the trigger conditions. But it can also serve to bolster situations that are covered by insurance but where bigger or quicker pay-outs are needed.

When a pandemic strikes that falls within the terms of the insurance window, emergency financing will be released from the facility based on a series of triggers that include the number of deaths, the speed of spread of the disease and whether the outbreak crosses international borders – all determined on the basis of data reported in the Disease Outbreak News bulletins, known as Dons, that the WHO issues periodically during an outbreak.

The parametric triggers for the facility – which are fixed for the life of the bonds – are specific to each disease that is covered and reflect the nature of those diseases. While the financial assistance provided by the PEF will only be paid to the 77 IDA countries, the triggers take into account cases anywhere in the group of middle-income countries that are eligible for assistance from the IBRD, adding a further 90-odd territories. In addition, pandemic flu was deemed so contagious that the terms for that disease take into account cases originating anywhere in the world.

The design of the facility and its global scope raise challenges in sourcing the appropriate data during an event to act as the inputs for the triggers. Partnering with the WHO is seen as the best option because of the quality and breadth of its data, its positioning with member countries and the formalization of its processes, where it is considered to be one of the most developed of international organisations.

To get to this point, however, the World Bank needed a model. A request for proposals was put out, with AIR Worldwide chosen over a handful of other modellers. The company has been around since the 1980s, is well-known in the field and already had a pandemic model. It had clients among the potential investor base and it had worked with the World Bank before.

While it's no surprise that the risk modelling is complex, the scale of what is involved is nonetheless startling. The AIR Pandemic Model, which the company updated in 2016 to add six new pathogens – including three that the World Bank was looking to tackle through the PEF, as the model already covered the others – simulates 1 million different outbreaks of pathogens, viruses and bacteria, across 24,000 municipalities. Not all the situations are pandemics, but they are all events that could severely impact a population. An event lasts on average 15 to 18 months and AIR models it day by day. All of this is mapped onto seven age groups and two genders. Outputs include mortality, morbidity and economic impact.

"We model the elements of a pandemic across the whole world," Doug Fullam, senior manager of life and health modelling at AIR Worldwide, tells Euromoney. "If you want to know the impact of a pandemic on young adults in the Democratic Republic of Congo, you have to be able to model them as opposed to people in the UK and the US. But to understand the risk in a US and UK pandemic, we also need to understand that risk in every part of the world – where it is starting and the spread pattern."

Historic data on previous incidents is fed into the model, but the crucial part is to control for how differently a historic event would affect today's more advanced and more interconnected world. This is where probabilistic modelling of the type that AIR engages in can help to fill in the gaps that result from simply extrapolating from infrequent historic data. AIR factors in epidemiological research into how pathogens spread, but also the current effectiveness of the healthcare infrastructure in the regions it models, considering the number of doctors per capita, or the availability of antibiotics.

All of this was brought to bear on the modelling for the World Bank deal's parametric triggers. "You could say that the basis risk is that you might be thinking about whether a historic event may or may not have paid out on this transaction, but the next event is likely to be very different, so you have to be careful not to design the parametric triggers to be too specific," adds Fullam.

For Michael Bennett, head of derivatives and structured finance in the treasury department of the World Bank, work on the PEF and the eventual bond issue started almost immediately in the wake of the Ebola outbreak in 2014 – and was going in earnest by early 2015.

"Our objective was to design something that would cover those diseases that were most likely to cause a serious pandemic in developing countries," he tells Euromoney. "We backtested to previous events, looking at when we would have wanted money and how much we would have wanted at that point in time." The findings from that helped to inform the coverage parameters of the PEF and the amount of coverage that the Bank was looking for.

Swiss Re and Munich Re worked with the World Bank on structuring, Swiss Re Capital Markets led the deal as sole bookrunner, while GC Securities – an arm of Guy Carpenter – joined the syndicate for the bond issue as a co-manager alongside Munich Re and as joint arranger of the swaps.

The bonds are issued by the IBRD and fall within its capital-at-risk programme, given that principal is at risk in the event of a pandemic triggering a pay-out under the facility. Alongside the $320 million of bond financing are $105 million of derivatives that transfer pandemic risk to counterparties. Exposure to the IBRD is nothing new for ILS investors, given that previous transactions that used a special-purpose vehicle structure would often see the SPV buy IBRD bonds as collateral, alongside treasury securities and money market funds.

The bonds are structured into two classes that cover different risks. The $225 million of Class A notes cover pandemic flu and coronavirus, while the $95 million of Class Bs cover coronavirus, filovirus, Lassa haemorrhagic fever, Rift Valley fever and Crimean Congo haemorrhagic fever. Donors Germany and Japan pay premiums into a PEF fund, and then a funding agreement between the PEF and the IBRD effectively results in them being passed through as bond coupons.

The bonds, which were sold as the first 144a transaction off the IBRD's capital-at-risk programme, have an initial term of three years but can be extended monthly up to a further 12 months.

Bennett says that while the deal ended up twice covered, there had been no assumption during the planning phase that it would be plain sailing. "One concern was that we were bringing brand new risk to the market in size. We were thinking of a minimum of $200 million on the Class A and $150 million on the Class B. It looked like a sizeable amount of risk to transfer."

That size lay behind the decision to opt for a combination of bond market financing and derivatives. Despite the pedigree of the IBRD as issuer and the World Bank backing the entire scheme, there were still always going to be limitations on what the bond markets might support, not least because the mandates of some cat bond investors restrict them to natural disaster risk. A simultaneous offering of exposure to the same risk through the derivatives market would help to bring in insurance and reinsurance parties who might prefer to take exposure that way.

Selling the same exposure to both markets is not unusual – although executing the two parts simultaneously with just the one book of demand certainly is. As well as ensuring broader distribution, the idea behind tapping both sectors is to create some pricing tension: some issuers prefer to tap the ILS market first to establish a benchmark for pricing and then approach the more traditional reinsurance market. In the case of the World Bank deal, bankers say that the two communities in fact reacted in broadly similar ways.

Guidance on the Class A tranche was a risk margin of 7.25% to 8%, while it was 12.25% to 13% on the Class B, a tranche whose risk profile was almost twice the Class A. The annualized attachment probability – in other words, the likelihood of a pay-out being triggered –  was 4.92% for the Class As and 9.44% for the Class Bs. The annualized expected loss – a function of the single-loss expectancy and the annualized rate of occurrence of a trigger event – was 3.57% for the Class As and 7.74% for the Class Bs.

"There are not really any perfect comparables in the market," says Bennett. "There is pandemic risk out there with some of the extreme mortality deals, but they are really aimed at insurance companies that are looking to reinsure tail events. Here we were talking about bringing about an intervention to contain the risk during the event, so there would be a higher expected loss."

The lack of meaningful comparables and the unusual nature of the risk and the pay-out structure meant that the deal took some explaining – and time was allowed for investors to do their own modelling and research. The simultaneous derivatives deal also needed to allow for negotiation of swap documentation. The book was held open for four weeks, closing in late June.

As orders flowed in – eventually hitting $850 million for the $425 million of risk – it was clear that better terms would be possible. Several rounds of tightening followed before final pricing ended up at a risk margin of 6.9% for the Class A and 11.5% for the Class B, equivalent to six-month Libor plus 6.5% and 11.1%, respectively.

"We were pleased by the level of interest and how quickly the deal built," says Bennett. "But by then we had done our due diligence and extensive premarketing, so we had good confidence going in that we would be able to get the deal done at good pricing."

There were 26 buyers of the two classes of bonds – with several names putting in orders for more than $50 million – and six counterparties for the concurrent swaps. The Class A notes were overwhelmingly bought by dedicated catastrophe bond investors, who took 62%. Asset managers took 20%, pension funds 14%. The Class Bs saw a very different profile, with pension funds taking the largest portion, at 42%. Cat bond funds took just 35%. In both tranches, the bulk of the notes were sold into Europe.

To the casual observer, pension funds buying the largest slug of the riskiest tranche of a capital-at-risk pandemic bond yielding more than 11% might sound like an accident waiting to happen. After all, there are triple-C rated corporates pricing deals more tightly than that right now.

The reality, argue those who work in this sector, is somewhat different. As the small number of investors in the book for the World Bank deal makes clear, these are specialist buyers. And while even five years ago catastrophe bond investors were often relying simply on external risk modelling, nowadays they are more likely to have their own models. "Investors have spent 20 years building up specialist knowledge and they hire managers who are experts in this space," says Cory Anger, global head of ILS origination and structuring at GC Securities. "They may have specialised ILS funds, and they look at this space as a low correlation with traditional asset classes."

Others agree. “The size of the ILS market in comparison to other asset classes is small, with $30 billion of outstanding catastrophe bonds versus $14 trillion of US Treasury bonds or $9 trillion of US corporate debt, and investors who are not specialised ILS investors are typically allocating only a small fraction of their total assets into ILS,” argues Andrew Fish, a catastrophe modelling associate director at Aon Securities, the leading insurance and reinsurance investment bank.

He also notes that despite the relatively small size of this fixed income market compared with the others, much if not all the novelty premium charged by investors has disappeared over the past five years or so: “The transparency of the market enables an efficient pricing of the risk for insurance-linked securities.” 

This notion of ILS being an uncorrelated diversifier is not quite as simple as it sounds, however. Changes in economic conditions could be expected to impact a country's ability to respond effectively to a pandemic – perhaps by restricting investment in local healthcare or through a greater prevalence of worsening housing and sanitary conditions, for instance. In addition, there has been no pandemic-related pay-out on an extreme mortality bond. The lack of loss history makes predictions hard to assess. All that said, the investments being made by traditional forms of investor such as pension funds are tiny relative to total assets under management, so the impact of losses is limited and well spread. Pension funds have perhaps $35 trillion under management globally, while the reinsurance market adds another $450 billion. A big pension fund might invest 1% or 2% into ILS or similar securities, at most. "This is like grains of rice to them, given overall assets under management in traditional capital market asset classes," adds Anger.

The interest shown by pension funds is in keeping with a broader trend, argues Jean-Louis Monnier, global co-head of ILS at Swiss Re. "After the financial crisis, we saw a real change in the composition of end-investors towards those with very long-term investment horizons, like pension funds," he says. "Whether they are investing directly or via specialist ILS funds, we see pension money grasp the long-term value of an investment in this asset class and be willing to accept the risk of potential losses."

For the more typical ILS and catastrophe investors – and for insurers – the deal's structure offers something of a hedge. Not only does it pay quite handsomely during its life – albeit with the chance of capital loss – but if you already have pandemic exposure, then the effects of the PEF can mitigate this to some extent because it is designed to reduce the seriousness of an event.

There is also good diversification to be had here for the dedicated community too. The bulk of ILS exposure is concentrated in natural perils in the US, such as earthquake and windstorm, meaning that correlation between the World Bank's deal and the broader ILS market is low.

Where could this all end up? The World Bank says it is working on other projects along similar lines, and Bennett leaves open the possibility of more deals relating to the PEF. But the intriguing prospect is the extent to which the idea of using parametric triggers to generate funds while an event is taking place can be borrowed from the pandemic bond deal for other types of issuer or other types of risk.

"You might be a healthcare network facing costs relating to pandemics, but there are also a lot of other industries that are affected as contagion is taking effect, like travel and tourism," says Anger at GC Securities. "Life reinsurers have always been more aimed at the back end of an event with extreme mortality cat bonds, i.e. surplus replenishment, but we are moving from an end-result approach to in-the-moment type of coverage."

Bankers agree that trends in developed markets could lead to greater concerns over time relating to diseases, especially if vaccination rates in some areas continue to fall perilously close to minimum thresholds for effectiveness – or even below – for conditions such as measles. But they argue that this has had no effect on pricing so far in the life market. "This isn't seen as something that is affecting your 50-something life insurance policyholder – it may be a factor in the future," says one. There is also the possibility of other companies borrowing the same structure as the World Bank deal to secure pandemic coverage for themselves through the capital markets.

"Because pay-out from a parametric trigger is quick, efficient and easy to determine, parametric products [such as the IBRD notes] are appropriate for disaster-relief purposes," says Fish at Aon. "As such, this transaction may set a precedent for other public-sector companies to use similar products in lieu of traditional disaster insurance policies.”

Issuers could also tailor mortality statistics to match their underlying risks precisely, rather than relying on public data such as from CDC. Imagination seems to be the only limit to what kind of risk might be offset over time. One structurer speculates that cyber-risk bonds could be on the agenda before too long – recent high-profile virus attacks have obvious similarities with pandemics. 

Whatever the chances of that, it seems certain that risk modellers such as AIR Worldwide have a busy future. Their market is only set to get bigger.