21 October 2014ILS

A 100 billion dollar question

In today’s reinsurance market, the question that often comes up is: what impact would a significant catastrophe loss have on risk transfer? With respect to the catastrophe bond market in particular, many would expect a large industry loss event to cause high levels of principal reduction, but in many cases this may not be so.

Leveraging the AIR US hurricane model, the AIR US earthquake model, and AIR’s comprehensive catastrophe bond database (which contains all active classes of catastrophe bonds in the market and represents more than $20.6 billion in principal outstanding), we can assess the relationship between industry loss and catastrophe bond principal loss to investors.

Let’s take a look at three hurricane events from the AIR US hurricane model’s 10,000-year standard catalogue and consider them in the context of a ‘market portfolio’ consisting of all catastrophe bonds currently on risk, weighted by size of issuance.

Event 1

This makes landfall in Palm Beach, Florida, as a Category 4 hurricane, heads west across Florida, and enters the Gulf of Mexico before making a second landfall near New Orleans, Louisiana, once again as a Category 4. The event causes $104 billion of modelled insured industry loss, which corresponds to a 1.40 percent US hurricane occurrence exceedance probability (OEP), or a one-in-71-year loss.

To put this loss in context, it would be the second highest industry loss-causing hurricane in history, after the 1926 Miami hurricane (~$119 billion in modelled industry loss with today’s exposures). While this event would cause principal loss to 35 securities, it erodes only 9.7 percent of the invested principal in the market portfolio. Certainly it would be a dramatic disturbance to the cat bond market, but the principal loss is perhaps smaller than many would expect for a more than $100 billion event.

Event 2

A much smaller industry loss event could have a similar impact on the cat bond market, simply because it impacts a different geographical area with higher catastrophe bond exposure concentrations.

Event 2 makes landfall as a Category 2 hurricane in Nassau County, New York, and causes industry insured losses of $36 billion (6.37 percent OEP or a one-in-16-year loss). It causes principal loss to just 14 securities but, similar to the much larger Event 1, is estimated to cause a principal loss of 10.7 percent to the market portfolio.

"The analysis demonstrates that industry exposure from a catastrophic event is not the same as catastrophe bond exposure to the same catastrophe."

This comparison illustrates how two hurricanes at different points on the OEP curve could have similar impacts on principal reduction. However, the purpose of this comparison is not to illustrate that $100 billion hurricane events incur relatively small principal loss, but rather that it is possible for two events with significantly different industry losses to cause a very similar principal loss to the portfolio.

Event 3

This makes landfall in Charleston, South Carolina, as a Category 4 hurricane, then continues to impact major metropolitan areas along the Eastern Seaboard, eventually making its way up to Queens, New York, as a Category 2 hurricane. The event causes an industry loss ($96 billion, 1.61 percent OEP or a one-in-62-year loss) similar to Event 1, but it incurs a significantly higher principal loss. This event causes principal loss to 71 securities and a 41.5 percent principal loss to the cat bond market.

From these three simulated hurricane events, it is clear that high variability is possible in the relationship between principal loss to the cat bond market and industry losses. There are 41 events in AIR’s 10,000-year standard hurricane catalogue that cause industry loss between $90 and 110 billion (approximately 21,000 events are simulated in the catalogue). These 41 events cause principal loss to the cat bond market ranging from 9 percent to 42 percent, with an average of 20 percent.

It must be noted that, for the purposes of this analysis, Events 1, 2, and 3 are all qualifying events.

What about earthquake?

Earthquakes are the second largest risk to the catastrophe bond market, with 48 percent of the principal in the market portfolio exposed to the peril, but do $100 billion earthquakes exhibit the same variability in principal loss as hurricanes?

There are 10 events that cause industry loss between $90 and 110 billion. These events have principal reduction ranging between 15 percent and 24 percent to the market portfolio, which is a significantly narrower range of portfolio impact than we observe for hurricane.

Table 1 below illustrates the relationship between size of industry loss and principal reduction to the market portfolio for earthquake and hurricane. While the mean and median principal reductions for these two perils are comparable, the variability, as quantified in the standard deviation, is significantly higher for US hurricane than it is for US earthquake. The lower variability in principal reduction for US earthquake, as compared to US hurricane, can be attributed to several factors, including the concentration of large events near known faults and less geographical spread of losses from a single event.

The analysis demonstrates that industry exposure from a catastrophic event is not the same as catastrophe bond exposure to the same catastrophe. Principal loss to the cat bond market is a function not only of industry exposure but also the concentration of principal in different region/peril combinations of each bond in the market, pre- and post-landfall activity of each event, cat bond structures (eg, aggregate versus occurrence, or indemnity versus industry loss), the sequence of events within a risk period, and exposure concentrations.

The analyses were performed using the entire catastrophe bond market portfolio and with one question in mind; however, a skilful portfolio manager or underwriter can optimise their exposure to varying tolerances of risk by conducting multidimensional analyses for the market as a whole or for their own positions. Further, by studying a full distribution of stochastic scenarios, rather than a few select events, more robust analytics can be performed which can aid in strong risk management.

Rhodri Lane is manager, insurance-linked securities at AIR Worldwide. He can be contacted at:

Adil Imani is a risk analyst at AIR Worldwide. He can be contacted at: