Ratings agency Standard & Poor’s has assigned an investment grade rating to a catastrophe bond for the first time since 2008, as the remote risk nature of the underlying Japanese earthquake risk helped the Kizuna Re II 2015-1 cat bond to a ‘BBB- (sf)’ preliminary rating assessment.
The rating on a cat bond is based on the lowest factor out of the natural catastrophe risk, the collateral asset rating and the ceding insurers rating. With the Japanese earthquake exposure in the Kizuna Re II 2015-1 cat bond being a remote layer of sponsor Tokio Marine’s reinsurance programme it is the lowest rated factor, but being so remote it is deserving of an investment grade preliminary rating of ‘BBB- (sf)’.
S&P explained; “This is the first property related nat-cat bond we’ve assigned an investment grade rating to since 2008.”
Gary Martucci, Director of Financial Services Ratings at S&P, explained some the reasoning behind the rating to Artemis; “TMNF’s market share offset some of the stress typically applied to a transaction where the cedent has a much smaller market share since their losses will more likely mirror the industry. In addition, given TMNF’s market share, by looking to the highest probability of attachment based on the variable reset, we believe there’s a significant amount of conservativism captured.”
The last investment grade catastrophe bonds rated by S&P were the mid-2008 issue Vega Capital Ltd. (Series 2008-1) and a novel equity and bond type catastrophe securitisation Globe Re Ltd. from the same year.
The notes being offered in the Kizuna Re II Ltd. (Series 2015-1) catastrophe bond will provide Tokio Marine & Nichido Fire with a collateralized source of reinsurance protection against losses from Japanese earthquake risks, including fire following, on an indemnity and per-occurrence basis.
Also included are any losses resulting from tsunamis or volcanic eruptions that are caused by an earthquake during the term of the cat bond and S&P noted that the volcanic eruption exposure is unmodelled. There’s an element of earthquake triggered flood risk in this cat bond too, with flood losses, including any caused by dam or levee rupture or sprinkler leakage, covered.
Martucci went into more detail on the preliminary rating of these notes, saying; “To assign the nat-cat risk factor, we applied an adjustment to the time-dependent and time-independent results to reflect the possibility that the probability of attachment may be greater than the models had anticipated.
“Typically, we use the more conservative of these results, but there was a large difference between the two in this particular transaction. In year 1, there was more than a 10-fold increase in the estimated probability of attachment by moving from time-dependent to time-independent modeling, 2.1 bps versus 21.4 bps. This large difference seems to result at least in part from the recent Tohoku earthquake. The time-dependent result is lower because such an event decreases the amount of stress on a fault, making subsequent events less likely.”
An interesting point on this deal from S&P’s pre-sale report is the fact that the underlying portfolio can change, effectively raising the expected loss, but the coupon will remain the same.
S&P explains; “This issuance is structured to keep the attachment and exhaustion points constant, so, as exposures change, the probability of attachment, exhaustion, and the modeled loss will increase and decrease, subject to a maximum expected loss of seven basis points (bps). The interest spread will remain constant.”
However with this being such a remote risk transaction and with the deal likely to have a huge multiple investors are still being compensated for this additional uncertainty while the sponsor will benefit from additional flexibility in the structure as its portfolio changes.
S&P explains further:
AIR did the modeling using TMNF’s exposures as of March 31, 2014, and AIR’s time-dependent catalog. The probability of attachment, expected loss, and probability of exhaustion based on this catalog were 0.021%, 0.018%, and 0.016%, respectively.
When rating catastrophe bonds linked to earthquakes, we look at both time-dependent and time-independent analyses, and usually use the more conservative of the two. The time-independent analysis probability of attachment is 0.214%. The time-independent expected loss and exhaustion probabilities are 0.194% and 0.181%, respectively.
This issuance is structured to keep the attachment and exhaustion points constant, so, as exposures change, the probability of attachment, exhaustion, and the modeled loss will increase and decrease, subject to a maximum expected loss of seven basis points (bps). The interest spread will remain constant.
When rating nat-cat bonds that permit variable resets, we look to the highest probability of attachment permitted, which is based on the modeled annual expected loss cap of seven basis points (bps). Based on the time-dependent occurrence exceedance probability (OEP) curve, if the expected loss is seven bps, the attachment point would be ¥238.6 billion and the probability of attachment 10.5 bps. Based on the time-independent catalog, the probability of attachment and expected loss would be 35.5 bps and 30.5 bps, respectively. Our analysis for the second through fourth risk periods assumes the expected loss for each reset to be seven bps. We then stress the results to determine the nat-cat risk factor.
To assign the nat-cat risk factor, we applied an adjustment to the time-dependent and time-independent results to reflect the possibility that the probability of attachment may be greater than the models had anticipated. Typically, we use the more conservative of these results, but there was a large difference between the two in this particular transaction. In year 1, there was more than a 10-fold increase in the estimated probability of attachment (POA) by moving from time-dependent to time-independent modeling (2.1 bps versus 21.4 bps). This large difference seems to result at least in part from the recent Tohoku earthquake. The time-dependent result is lower because such an event decreases the amount of stress on a fault, making subsequent events less likely.
The stress factor that we applied to determine nat-cat risk factor reflected a number of factors. Our criteria list indicative stress levels for varying transactions with industry loss transactions having a 10% stress and indemnity transactions a 20% stress. Although losses for this transaction are calculated on an indemnity basis, TMNF’s large market share in some ways makes it more similar to an industry loss transaction. We would not expect a carrier of its size to have losses that greatly differ from its proportional share of the market’s losses. We also considered that the transaction contains some small amount of unmodeled risks and the transaction had a variable reset feature that we were treating very conservatively by looking to the highest probability.
We then select the next rating category below this adjusted POA that’s greater than or equal to the adjusted probability of attachment from our nat-cat risk factor table. This adjustment results in varying margins between the initial parameter values needed to trigger a payment from the noteholders to the issuer under the reinsurance agreement and the parameter values at the probability of attachment commensurate with the assigned rating.
There is a time element to qualifying losses under the terms of this cat bond as well. S&P said:
Losses from any event must be within a period of 168 consecutive hours and selected at TMNF’s discretion–provided that this 168-hour period does not begin earlier than the date and time of the first recorded individual loss to TMNF from that earthquake. The epicenter does not have to be within the covered area. The covered area is the territory of Japan (all 47 prefectures).
Also of note S&P highlights, as usual, some positive and negative factors on the deal. On the positive side S&P says that Tokio Marine & Nichido Fire will retain 10% of every loss. On the negative side the rating agency note that commercial exposures account for more than 82% of the expected losses for the cat bond and that 34% of building construction types are coded as unknown for earthquake policy net limit of liability.
Mitigating factors of interest include the fact that in the commercial/industrial line of business, occupancy is known for 99.8% of the total insured value. Meanwhile any losses from fire following an earthquake are limited to 5% of the fire total insured value with a cap of ¥3 million per occurrence and per premises (for factories, this cap is ¥20 million per occurrence and per premises).
On the risk modelling side, since 1700, there have been two events–the 1923 Taisho Kanto quake with the epicenter in the Kanagawa prefecture (part of the greater Tokyo area) and the 1707 Hoei earthquake with the epicenter in the Wakayama prefecture–that would have triggered the bond.
S&P explains; “The Hoei was the most powerful earthquake in Japan before 2011’s Tohoku earthquake, and it possibly caused an eruption of Mt. Fuji later that year. This 1923 event as modeled resulted in a 100% principal reduction. The modeled principal reductions from the 1707 event were modeled with (73% principle reduction) and without (68% principal reduction) the effect from the tsunami.”
Importantly, the number of historical earthquake events that could have impacted this cat bond does not increase if the 0.07% increased expected loss figure is used.
It’s also worth noting that the 2011 Tohoku earthquake would not have come anywhere near attaching the Kizuna Re II 2015-1 cat bond notes, as its ultimate net loss would only have been ¥95 billion and the attachment point is ¥310 billion.
On the subject business exposure covered by this cat bond, S&P explains:
The covered exposures comprise commercial business that includes, but is not limited to, fire policies covering buildings used as shops and offices, as well as their contents; and industrial business that includes fire policies covering properties such as factories and warehouses, as well as their contents. Although the majority of policies fall within the category of fire policies, these two lines comprise policies categorized as fire, engineering, movable all risks, and business interruption policies (less than 30 policies have business interruption coverage) unless otherwise specified. Marine policies are not part of the covered exposures. Earthquake coverage for both property damage and business interruption is an endorsement to the standard contract earthquake endorsements that cover debris removal expenses (limited to 10% of the earthquake payout amount). The coverage includes fire policies and engineering and movable all risks policies. In all fire policies, earthquake fire expense insurance (EFEI) is covered in the standard contract. The modeled results indicate the majority of expected losses come from earthshake.
Personal business consists of fire policies written on buildings used only for dwellings and household goods. Policy holders purchase earthquake coverage as a separate endorsement.
For personal lines, all dwelling earthquake covers except for EFEI are ceded to Japan Earthquake Reinsurance Co. Ltd., the government-backed plan, and, therefore, are not reinsured under the reinsurance agreement. The covered exposures include policies in which TMNF is not the lead co-insurer. This accounts for less than 10% of earthquake total limit.
The overwhelming contribution to expected loss is related to the commercial and industrial policies. The total insured value for commercial, industrial, and personal business fire policies are based on replacement value or actual cash value (estimated replacement value less depreciation of property). The majority of the policies are written on a replacement-value basis.
As is typical Tokyo contributes the greatest amount to the expected loss for the Kizuna Re II 2015-1 cat bond, at 14.6%, with Shizuoka next at 12% and Kanagawa at 11%.
As ever S&P demonstrates the value in a rating report with the thorough nature of its analysis of the catastrophe bond deal.
The Kizuna Re II Ltd. (Series 2015-1) cat bond is in the market right now and the expectation is that it will close and settle in the last week of March. Read all about the transaction in the Artemis Deal Directory.