The Guy Carpenter Hurricane Index helps (re)insurers assess the potential losses from hurricanes making landfall in the Atlantic Basin by defined regions: Northeast, Southeast, Florida, and Gulf. Guy Carpenter created this index in response to the (re)insurance industry’s questions about the credibility of catastrophe models after the 2004 and 2005 hurricane seasons and their need to better understand and manage their catastrophe risk. Unlike most indices, which do not differentiate by location and only take into account frequency or severity and historical climatologic information, the Guy Carpenter Hurricane Index also considers current meteorological information and forecasts about live events, the forecast landfall location, and the forecast landfall probability and the conditional severity.
These indices can be used to purchase reinsurance prior to the inception of or during a named event.
The GC ForeCatTM Indexed Industry Occurrence Exceedance Probability Curve (OEP), which is based on a seasonal view of regional landfall propensities, drives the selection of coverage prior to a named storm. For each of the regions covered, the GC ForeCat Indexed Industry OEP Curve is defined as:
GC ForeCat Adjusted Reference Vendor Model Industry OEP / Reference Vendor Model Expected Loss given a storm
The following represents the GC ForeCat OEP curve and GC ForeCat Indexed Industry OEP curve for the Gulf for 2008.
The GC ForeCat rate for the Gulf in 2008 was 1.35, and as the climatologic rate for the Gulf was 0.66, this implies that GC ForeCat was predicting twice the storm activity as that implied by the historical average. The above chart illustrates that GC ForeCat predicted more events making landfall in the Gulf, which also yields a higher OEP curve than the vendor model. Thus, an insurer may want to purchase additional frequency or vertical protection in the Gulf region. This can be done on an ultimate net loss or parametric basis for the entire region - or for specified gates within the region.
On a parametric basis, for example, after reviewing the 2008 Gulf GC ForeCat OEP relative to the vendor model OEP, an insurer may be concerned about the seasonal 1-in-100 year loss of USD38 billion — compared to the near term view of USD30.75 billion. A cover satisfying this concern would attach at an index value of 10.4. Coverage could be pro rata - i.e., based on a window (exhaustion point), or binary. Thus, if protection up to a 1-in-200 year loss event is sought, the exhaustion index value would be 14.2. Coverage is triggered upon a named event yielding an index value exceeding 10.4.
Index values are updated when there is a named event in the Atlantic Basin. There is a separate GC LiveCat Hurricane Index for each region, and it is updated every 12 hours. For regions where there is no landfall, the Index Value remains at 1.0.
The updated Index Value, GC LiveCat Hurricane Index, is:
The conditional expected loss given the forecast of named event/the near term vendor model expected loss given a storm.
Thus, if an insurer purchased the aforementioned coverage for Hurricane Ike in 2008, there would be no payment made as the landfall Index Value equaled 2.27 (as seen below). As Ike was approximately a 1-in-17 year return storm on the vendor model distribution, for a cover to have a payoff, it would have had to attach at a value less than 2.27.
Coverage also could be purchased on a “live” basis, when a storm is present in the Atlantic Hurricane basin. As stated above, we define the GC LiveCat Hurricane Index as:
The conditional expected loss given the forecast of named event / the near-term vendor model expected loss given a storm
Prior to the inception of a live storm, the initial value of the index is always equal to 1.0 Guy Carpenter will be publishing GC LiveCat Index, as well as its distribution, in order for parties to be able to determine the structure. Similar to the pre-season product, determination of whether a payment is made is predicated upon the GC LiveCat Index value at landfall. Index values and distributions within 24 hours of landfall will only be distributed to licensed/transacting parties.
For example, on September 9, 2008, an insured may have decided to purchase coverage with an attachment point of 1.5 and exhaust at a value of 3.0. Given the GC LiveCat Landfall Index was equal to 2.27, the cover would have been triggered.
Loss settlement can be paid in two different ways: modeled loss or selected limit on a pro rata basis or binary basis.
For example if the insured wants to purchase USD100 million of coverage on a binary basis, the payment is USD100 million. If purchased on a pro rata basis, the loss payment is (2.27-1.5)/(3-1.5)*USD100 million - or USD51.3 million.
Guy Carpenter also calculates modeled loss estimates every 12 hours during the storm using the WSI LiveCat Forecast and a defined portfolio. For Ike, at landfall, Guy Carpenter calculated an industry modeled loss estimate for the Gulf region of USD6.73 billion. For a modeled loss cover, the contract will have a stated dollar attachment point and limit. Thus, if the attachment point is USD6 billion and the limit is USD100 million, then the loss settlement is USD100 million. Given that the Guy Carpenter modeled loss estimate does not include estimates for flood, a factor can be included in the contract to adjust for this.
The GC Hurricane LiveCat Index provides parties with the ability to purchase live catastrophe coverage up to five days before the National Hurricane Center forecast. Further, unlike the PCS index, it gives the parties information about the event rather than just a severity value based on historical information and post-event loss statistics from the industry. In addition, loss settlement is finalized a few days after landfall as opposed to 12 to 18 months after for a cover based upon PCS loss settlement. GC ForeCat and GC LiveCat provide more information to allow (re)insurers to buy backup coverage if required - not to mention enhanced, optimal, and efficient coverage as coverage can be defined on a county, gate(s), storm intensity, or regional basis.