Golden State Re II Ltd. (Series 2014-1) – Full details:
This transaction is very similar to the SCIF’s 2011 Golden State Re Ltd. deal.
A new special purpose vehicle has been established for this renewal issue, with Golden State Re II Ltd. being incorporated in Bermuda as a special purpose insurer.
This Golden State Re II cat bond issuance will see a single tranche of Series 2014-1 notes marketed to investors for the sponsor SCIF.
The deal is expected to be sized at $150m or over, depending on ILS investor appetite.
The SCIF will enter into a reinsurance agreement with Golden State Re II, in order to benefit from the source of fully-collateralized reinsurance protection.
The protection will be afforded on a per-occurrence basis using a modelled loss trigger in the same fashion as the 2011 cat bond issue. The transaction will cover the SCIF until January 2019, so more than four years which is a year longer than its 2011 issues tenure.
The covered area is for earthquake events in the entire U.S., but as with the 2011 deal as much as 99.99% of the SCIF’s insurance portfolio is focused on California, as its name gives away. As a result the risk is focused on California area earthquakes.
The modelled loss trigger uses a variety of inputs and a calculation process to derive whether an event has triggered the cat bond. The trigger is of similar construct to the 2011 deal, using the exposures of a notional portfolio of workers compensation risks in the SCIF’s portfolio, earthquake severity factors (ground motion), geographic distribution of the covered portfolio, types of buildings covered, time of day and the day of week an event occurs as some of the weighting factors.
After an event, which we believe has to be an earthquake of magnitude 5.5 or greater, losses will be modelled deterministically, so not related to actual injuries and fatalities, using the earthquake event parameters and this will be modelled against the notional portfolio using day/time weighting to determine an index value and notional modelled loss amount.
A calculation process will be run post-event to derive an index value which will be compared against an attachment point. The attachment point is at 1,000, and the exhaustion at 1,903. The 2011 deal also attached at 1,000 but exhausted at 1,447, so this deal may cover a larger layer of the SCIF’s reinsurance programme.
The initial attachment probability for the notes will be 0.5%, while the initial exhaustion probability will be 0.11% and initial expected loss 0.25%. The expected loss of the 2011 deal was 0.36% for comparison, which suggests the 2014 Golden State Re II cat bond is a little lower risk.
As a result the coupon guidance is low for this cat bond, although low is to be expected in the current market environment anyway. The deal is marketing with coupon guidance of 2.2% to 2.7%. Again for comparison, the coupon of the 2011 Golden State Re cat bond was 3.77%.
Standard & Poor’s gave the notes a preliminary rating of ‘BB+(sf)’, noting that:
The modeling results assume events are evenly distributed throughout a day. Since the timing of an event will have an impact on the loss amount, we reviewed two sensitivity tests that adjusted (in hourly increments) the time of each stochastic event by plus/minus two hours and six hours from its original time. Using the initial index values and the adjusted index values, the highest notional modeled loss from each event was used to construct a new exceedance probability curve and the results were consistent with a nat-cat factor of ‘bb+’.
The rating is based on the lowest of the natural catastrophe risk factor, ‘BB+’, the rating on the assets in the reinsurance trust account, ‘AAAm’, and the creditworthiness of the ceding insurer.
The ‘bb+’ natural catastrophe risk factor was derived by S&P using the risk model that the transaction was structured with. RMS is the risk modelling firm in question and its RMS U.S. Earthquake Casualty Model, updated in 2014, as implemented in RiskLink version 13.1 is the model in use for the deal.
S&P notes that the risk modelling for the cat bond is based on the notional portfolio of SCIF risk as at 30th April 2014, so if the cat bond is triggered the actual exposure will be different due to the passage of time, but not sufficiently to cause any concern. The notional portfolio is expected to adequately reflect the covered business throughout the life of the deal. An updated notional portfolio may be provided with the reset, which should help to keep any gap between notional and actual exposure to a minimum.
S&P always tests the modelling results and sometimes applies incremental stress factors, such as adjusting the modelled probability of attachment, if its criteria require it. S&P’s catastrophe bond rating criteria prescribes that it applies a 7.5% stress level to a modelled loss transaction, although this can differ depending on the company’s strengths, concerns and mitigating factors.
S&P noted its concerns about the transaction. Firstly, the time of day factor having such a potential impact on the resulting event index level (hence the additional testing mentioned above. Secondly, the fact that there may be triggering events which are not captured by the model, always an issue with an earthquake bond particularly in an area with so many faults as California. And third, the investment risk that the Treasury money market funds used as collateral may have, although as we all know these are about as low-risk as collateral assets can get.
Based on RMS’s risk analysis there have been three historical events which could have generated an index value high enough to breach the attachment level of the notes had they occurred at specific times of day. These are the 1857 Fort Tejon earthquake had it occurred at 2 p.m. and the 1906 San Francisco and 1994 Northridge earthquakes, had they occurred between 9 a.m. and 3 p.m.
However, each of these historical earthquake events actually occurred at times where the modelled index values that RMS generated were not in excess of the attachment point. This shows just how the modelled loss trigger has been constructed to closely match the modelled loss with a workers compensation claim level, thus providing the cover the SCIF seeks.
The main exposure for this earthquake is in California, despite it covering all 50 U.S. states. In fact 99.99% of the SCIF portfolio is located in California, hence the coverage is targeted there. For an event to qualify it must be at least magnitude 5.5, which does mean that recent events such as the Napa earthquake would have qualified and may have resulted in the index values having to be generated by the calculation agent, although with no chance of reaching the trigger we’d imagine.
In terms of which part of California is most risky for this cat bond, Los Angeles County contributes 52.1% of the expected loss, Orange County 10%, Alameda County 6.1%, San Bernardino 5.8% while the rest contribute under 4%. Even San Francisco only contributes 3.6% to expected loss. This shows that the focus of the notional workers compensation portfolio’s exposure is on Los Angeles.
Employees working in offices make up 47% of the employees covered, which again ties in with a focus on LA. 9am to 4pm on a weekday is the riskiest time for this cat bond with the modelled contribution by time of day highest between these hours, according to S&P.
The way the cat bond trigger has been constructed ensures it provides the coverage the SCIF requires, for workers compensation claims it suffers due to earthquakes. The index value could therefore be very different for an earthquake occurring outside working hours, or at the weekend, compared to one which occurred in the middle of a working day.
The contribution to modelled losses by time of day even dips slightly around lunchtime, as more workers will be outside of the workplace should an earthquake occur at that time. This is intelligent use of modelling and data to construct a catastrophe bond that meets the needs of the ceding party.
The Golden State Re II catastrophe bond upsized by two-thirds to reach $250m in capacity, offering the SCIF $50m more than its maturing cat bond.
At the same time the price guidance narrowed and fell approximately 5% to a range of 2.2% to 2.45%.
At final pricing the coupons settled at the bottom end of the initial range, at 2.2%, which represents a -10% drop from the mid-point of guidance, or as much as -18.5% from the top of that initially marketed range. The deals size hasn’t changed and remains at $250m.