S&P highlights complexity in Aozora Re Japan typhoon catastrophe bond

More details have come to light on one of the recent catastrophe bond transactions to launch, the $100m Aozora Re Ltd. (Series 2014-1) Japanese typhoon cat bond from primary Japanese insurer sponsors Sompo Japan Nipponkoa Insurance.

Rating agency Standard & Poor’s has published its presale report and assigned the Aozora Re cat bonds two Series 2014-1 tranches of Class A and Class B notes a preliminary rating of ‘BB(sf)’. S&P notes that this is the first catastrophe bond it has rated that provides indemnity coverage for losses from typhoons in Japan.

S&P’s presale reports always include a number of strengths and potential concerns about any new catastrophe bond, but in this case the concerns are considered significant enough to warrant an adjustment to the probability of attachment during S&P’s assessment of the bond. This adjustment was greater than the 20% indicative stress level which is part of S&P’s rating criteria.

As we said in our first article on this deal, the Aozora Re 2014-1 cat bond will provide Sompo Japan Nipponkoa (SJNK) Insurance with a three year source of fully-collateralized reinsurance for qualifying Japanese typhoon losses, on an indemnity and per-occurrence basis. Two tranches of notes are being issued which have the same attachment and exhaustion points, with the only difference being the currency in which each class of notes are denominated.

The first interesting point to note from S&P’s presale report is that the Aozora Re cat bonds modelling was performed using SJNK’s exposures as at 30th September 2012. S&P notes that it has been told that the subject business has not changed materially since that date, however that is quite a long period during which subtle changes may have occurred.

S&P highlighted, as usual, a number of concerns with the cat bond transaction, however in this case they are a little more unusual than in other recent cat bonds to come to market.

S&P said that around 13% (based on total insured value) of the exposures covered had an unknown location, that around 34% of the initial modeled loss is from industrial and commercial lines and that demand surge is not modelled. These factors could result in unexpected losses to NKSJ, outside of the expected losses from the modelling we would assume. It also suggests that certain exposures have not been modelled fully given their location was unknown.

Mitigating factors noted by S&P include the fact that 66% of the expected loss contribution is from personal lines exposures and that there are no individual or groups of properties that create any concentration concern. Also that NKSJ will retain a percentage of the ultimate net losses for each covered event, thus reducing the potential impact to these cat bond notes.

Another potential concern noted by S&P is that the AIR risk model used for this cat bond does not contain historical events after 2006. However only one event after that date, Typhoon Roke in 2011, has caused losses in excess of ¥3.8 billion. Losses from Roke totaled approximately ¥29 billion, according to S&P.

Also, there have not been any typhoon events since 1947 that would have generated sufficient losses to reach the attachment point, said S&P. This is a remote risk, with a very high attachment point. 1959’s Typhoon Vera had the highest estimated ultimate net loss of ¥286.5 billion, however that number would need to increase by nearly 79% to reach the initial attachment level, and by 34% if the attachment was adjusted to the maximum under a variable reset. The next two events that generated the greatest ultimate net loss totals were 1961’s Typhoon Nancy and 1991’s Typhoon Mireille with loss amounts of ¥214.5 billion and ¥181.7 billion, respectively.

So, despite S&P’s concerns this cat bond remains a very remote risk. The concerns are some of the issues which has meant that Japanese sponsors have used parametric triggers in the past, with lack of data on exposures often cited as a reason that indemnity transactions took longer to take off in Japan. There remains no reliable industry loss index provider for Japan either, so for sponsors it is important to get investors comfortable with an indemnity trigger, despite concerns.

That might explain the remoteness of the risk in the Aozora Re cat bond. There may be an element of the sponsor seeking to get the capital markets comfortable with its risk management practices so that it can leverage ILS capacity again in future.

S&P notes the following on the complex exposures included in the subject business of the Aozora Re cat bond:

The subject business is a subset of the ceding insurer’s overall insurance portfolio and comprises fire insurance policies, movable-all risks (that is, property that can be moved outside of the insured premises) insurance policies, motor insurance policies, and engineering insurance policies. Motor insurance and fire insurance policies account for more than half of SJNK’s underwriting portfolio on a net premium income basis. All products cover typhoon and some products cover flood risks under standard policy forms. Losses related to movable-all risks, motor, and engineering insurance policies will only be covered under the reinsurance agreement by application of the adjustment factor.

Fire insurance is subclassed into personal, commercial, and industrial subject business lines of business. Personal consists of policies covering residential dwellings and contents. Commercial consists of commercial policies covering stores and offices, etc., as well as equipment and products within such premises. Industrial consists of industrial policies covering buildings and factories, as well as equipment, devices, products, and other unfinished products within such premises.

That is a complex mix of different property and risk types, more so than in many other catastrophe bonds. It also suggests an element of unmodelled risk, for example moveable risks are hard to model and devices and equipment in industrial facilities are again typically either difficult to confirm records of or unmodelled.

But this is quite typical of some of the more complex U.S. cat bonds to come to market these days, where there have been bonds with uncertain building engineering detail, factory contents, luxury yachts and offshore assets in recent years. On an indemnity basis investors need to buy into a sponsors internal risk management and underwriting processes to become comfortable with inclusion of such risks, as we have seen with bonds like AIG’s Tradewynd.

Also worth noting is that the region that contributes the most to expected losses is southern Japan, the area most prone to typhoon losses, with Fukuoka and Nagasaki prefectures contributing 19% of expected losses between them while Tokyo is just under 6%. Meanwhile wind contributes 90.8% of expected losses while flood only contributes 9.2%, on a building type basis dwellings are 41.3%, government housing loan scheme properties 24.6%, commercial 18.6% and industrial 15.5% of expected losses.

S&P notes that while the Class A notes are US dollar denominated the sponsor will be retaining any currency risk between the conversion from ultimate net losses, which will be calculated in JPY and any loss to noteholders which will be in USD.

The variable reset is worth noting as it allows the ceding insurer to adjust the modeled expected loss within a range of 0.52% to 1.04% at each reset date. S&P said that the probability of attachment based on the maximum expected loss is 1.15% and is the probability of attachment used to determine the nat-cat risk factor.

Interestingly, S&P also said that at the variable reset the expected loss can increase from the initial level 0.52% to 0.70% without any increase to the risk interest spread, so investors may not be compensated for some increase in risk profile with this cat bond, an unusual feature. Should the expected loss increase above 0.70% then the risk spread will be adjusted accordingly.

As a result of the first three concerns we mentioned above, the unknown location of some exposures, the industrial and commercial exposure and the lack of demand surge modelling, S&P said that it has adjusted the probability of attachment more than its rating criteria specifies to make allowance for these concerns and to derive its preliminary rating for the notes.

These concerns really come down to the added complexity of a Japanese indemnity bond with sponsors who are likely striving to improve data quality all the time. The Japanese market has some way to go before the quality of its exposure data is as good as seen in most U.S. catastrophe bonds, although it’s worth noting that some U.S. cat bonds do raise similar concerns among market participants.

The Aozora Re catastrophe bond is a little unusual, but as the first rated indemnity Japan typhoon bond that is perhaps to be expected. This is where the value of a rating becomes apparent, particularly from an investors point of view, as the extra due diligence performed by S&P will help to bring these concerns to light and enable investors to ask the right questions of the structurers at the bonds roadshow.

It is only be highlighting these concerns during the issuance process that investors can become better educated on the inclusion of more unusual or less well-modelled risks in cat bonds. It’s an important process for investors to go through, building their confidence in new sponsors and in the structurers work. Here S&P provides a very valuable service to the cat bond market.

You can read all about the Aozora Re Ltd. (Series 2014-1) catastrophe bond in our Artemis Deal Directory and we will update you as it progresses to market.