The impact of RMS v11 on ILS

When RMS revealed its 2011 US hurricane model release, Trading Risk’s headline trumpeted “Hurricane RMS” for its impact on the market. Before long, the new hurricane model, like a pop star diva, was known by just one name—v11.

New data and science, and vast increases in computational power, enabled RMS to improve the view of hurricane risk dramatically. With this enhanced view of risk also came an increase in the level of risk. For the industry overall, losses at the 100-year return period increased by around 50 percent. With the way the insurance-linked securities (ILS) market views risk, the changes seen were even larger, with expected loss (EL) on average doubling.

Insurance, reinsurance and ILS market participants braced for a reaction to this dramatic new view of risk. Would rates in the traditional reinsurance market harden significantly? Would catastrophe (cat) bond prices for hurricane bonds plummet? Would bonds originally modelled by RMS react differently from those modelled by AIR or EQECAT?

This article describes what we saw happen in the ILS market and provides a hypothesis for what is going on.

Sponsors ‘blown’ away

RMS had been communicating change guidance information for the best part of a year, so the market knew there were changes afoot. No sponsors wanted to issue a US hurricane cat bond using a model that was soon to change, nor to be the first to take a v11 bond to market. While some sponsors chose to use other models, many of the RMS shops chose not to issue cat bonds for the 2011 hurricane season at all.

The ILS market has traditionally been a ‘technically priced’ market. Spreads have been set as a relatively predictable multiple of the modelled EL on the bond. This worked well when all three modellers were relatively close together in risk estimates. However, with the RMS view of risk increasing so much, issuers (and the intermediaries that advised them) were unsure as to whether the market would maintain the same EL multiples, or whether it would keep the spreads roughly the same and adjust the EL multiples.

Rating agencies react

Immediately after the announcement of the increased probabilities of attachment, Standard & Poor’s (S&P) and A.M. Best reacted by asking the sponsors of all outstanding hurricane cat bonds originally modelled by RMS to provide updated ELs for their bonds. RMS worked with the sponsors to provide this information, which both agencies used to re-rate the bonds. In total, there were 11 bonds downgraded by S&P and one downgraded by A.M. Best.

What was interesting about this move was that S&P apparently also solicited AIR’s risk metrics for some of the bonds in question, to get a ‘blended’ view of risk for these bonds. However, RMS was not asked to provide risk metrics for any of the hurricane bonds outstanding where AIR was the modeller of record.

Certainly there is consistency in the rating agencies’ reliance only on the modeller of record for updated ELs to re-rate bonds. After all, there was only one EL considered when rating the bond originally. However, following this approach leads to an awkward situation where two identical bonds have different ratings because one was modelled by RMS and the other not.

New view, but no change in pricing

We have a window into investors’ reactions through the clients that license Miu, our ILS portfolio management platform.

“As with any significant advance in the science of catastrophe modelling, it can take many months to understand the changes and implement them into decision-making processes.”

We found that most of these investors were eager to get their hands on the new view of hurricane risk to assess the impact on their portfolios and to understand the RMS view of risk for new cat bond issuances. Investors were looking to incorporate the new RMS view of risk into their decision-making processes with urgency. Once investors were able to digest the new view of risk, they seemed broadly comfortable.

However, secondary market prices show that hurricane bonds have been changing hands at pre-v11 prices. Furthermore, the hurricane cat bonds that have come to market (using other modellers) were selling like hot cakes at spreads that reflected pre-v11 multiples of expected loss. So, if the pricing continues to follow v10-level pricing, and v11 is twice the EL, it must mean that the EL multiple for v11 must be cut in half.

Based on v10 measures of EL, the traditional market ‘rule of thumb’ was that:

Spread = (2 x EL) + 5 percent

Given that v11 ELs are roughly double what the other modelling firms are calculating, you would expect that new ELs should be:

Spread = (1 x EL) + 5 percent

When we plot v11 EL against spread at origination for recently issued cat bonds, we see almost exactly that (it shows Spread = (0.9 x EL) + 5.1 percent).

Having different EL multiples for the different modelling firms suggests that the market believes the models are good for assessing relative risk (Bond A is riskier than Bond B), but that they are using something else (intuition? conservatism?) for setting the absolute level of required spreads.

Waiting for the ‘trickle-down’

Insurers and reinsurers rely on models to understand both how much cover (reinsurance, retro or ILS) and the fair price for obtaining that cover. The market is still waiting for the primary insurers and reinsurers to fully implement the v11 view of risk into this process.

Common practice in the re/insurance market is to load cat model results to account for elements of risk that a company may view as outside the scope of the model. The v10 US hurricane model was commonly loaded, and the initial reaction from many sophisticated model users has been that v11 is more in line with their intuition, reducing the need for such loads.

Importantly, the reinsurance market has dealt with model differences for many years, with contracts commonly assessed using the results of two or three different models. Firms are therefore able to navigate the uncertainty with model-specific loadings—and under v11, probably significantly smaller RMS loadings than before.

As with any significant advance in the science of catastrophe modelling, it can take many months to understand the changes and implement them into decision-making processes. With more insurers and reinsurers using v11, we believe the firms that use RMS as their primary model will return to the ILS market—and will expect pricing that incorporates the new view of risk and recognises how it differs from other models.

As this happens we expect that the EL multiples could increase, reflecting the market’s view that the level of risk has increased; therefore the spread required for a given level of EL (by any of the modelling firms) should be higher.

New view of risk fits within uncertainty cushion

Our calculations suggest that if ELs are accurate, then historical spreads have had a healthy dose of conservatism—a ‘cushion’—built in, which could also be another reason why the market pricing has not changed following v11. This spread cushion was there to cover model uncertainty— exactly the type of uncertainty demonstrated by the change from v10 to v11, and the v11 view of risk still falls well within this.

Our required return model calculates the required spread for a catastrophe bond that is being added to a diversified portfolio of uncorrelated risks (eg, equities, corporate bonds). This model assumes that investors need to get paid for contribution to portfolio volatility, and they want to achieve a target Sharpe ratio of 2.0 (a suitably high ratio).

The model also assumes that the investors have only 2 percent of their total portfolio invested in any one peril. The chart below shows that for a bond with a 1 percent EL, investors should expect only 3.5 percent above Treasuries, and a bond with 2 percent EL should require a 6 percent spread.

Both of these required spreads are below the typical market spread of 7 percent (2 x 1 percent + 5 percent) that would be charged for a 1 percent EL.

So if the modelled risk is now 2 percent in v11 rather than 1 percent in v10, investors are still getting paid well for the risk they are taking.

Taking a leaf out of the re/insurer’s book, we could also view this as a change in the ‘loading’, giving us a new pricing paradigm of Spread = (1 x RMS EL) + 5 percent.

Time heals

We believe RMS v11 is a vastly superior view of risk and one that, while tough to adjust to in the short term, provides an appropriate benchmark for the management and transfer of hurricane risk.

As the re/insurance market continues its adoption of v11, we expect this new view of risk to carry on making its way into pricing and portfolio management decisions. In the meantime, even given the new view of risk and the current level of spreads, the ILS market represents good value to diversified fixed income investors.

Peter Nakada is managing director of RMS Risk Markets. For more details, please visit www.rms.com