Q&A with François Divet, Head of ILS (Part 2)

This is the second of two Q&As with François Divet, Head of ILS. You can read the first one here.

What attracted you personally to ILS?

ILS sits at the crossroads of (re)insurance and capital markets. With an actuarial background, I find that combination intellectually compelling. It requires both quantitative modelling and financial structuring expertise.

The market continues to innovate, with new perils such as cyber emerging. That evolution keeps the asset class dynamic and stimulating.

How much do you rely on models versus judgement?

We rely heavily on catastrophe models, both third-party and internally developed. They are essential for quantifying expected loss and tail risk.

However, we do not follow models blindly. Where we believe risks are under-modelled — for example US wildfire or other secondary perils such as convective storms — we apply additional loadings to reflect our own view of the risks.

Beyond modelling, structural features such as reset clauses and extension risk matter a great deal, and they require judgement. The quality of the sponsor of the cat bonds on the data provided and their capacity to accurately and in a timely manner report post-event losses are also fundamental for indemnity deals. So I would say it is a combination of quantitative modelling and qualitative assessment.

How do you mitigate downside if the models are wrong?

We mitigate downside risks in several ways. First, we impose hard exposure limits by region and peril, independent of model outputs. Second, we stress-test model assumptions by increasing the frequency and the severity of the events. Third, we may avoid risks that we consider insufficiently well-modelled or increase the required return to a level where it no longer makes sense to invest. We also diversify structurally — across layers, triggers and geographies. Even if two bonds reference the same peril, they may not respond identically to an event.

Finally, we monitor tail metrics such as 99% Value-at-Risk at portfolio level. Risk management is multi-layered; it does not rely on a single model output.

How do you incorporate climate change into pricing?

Some perils, such as earthquakes, are not climate-related. For hurricanes, climate change appears to affect severity more than frequency, with potentially more intense storms.

Catastrophe models are updated regularly to reflect evolving science and widening types of exposure. Attachment points — the level at which the bond’s principal is threatened — and structures also adjust over time. In my view, climate risk is largely absorbed through structural recalibration rather than through a simple widening of spreads.

So while risk levels evolve, the overall risk–return balance for investors can remain relatively stable.

How do you separate short-term weather volatility from long-term trends?

When we calculate expected loss, we rely on long-term stochastic simulations. These models incorporate structural climate trends but are not designed to trade short-term meteorological forecasts. We are cautious about short-term seasonal predictions — for example, unusually warm sea temperatures — because historically they have not always translated into realised outcomes. What is important is not the number of forecasted events, but where the events make landfall. A single event making landfall in Miami will have a greater impact than several events making landfall in a sparsely populated area. Our approach is anchored in long-term modelling rather than tactical weather views.

What happens to liquidity after major events?

After a major catastrophe, prices of potentially impacted bonds fall and bid–ask spreads widen. It can be more difficult to trade distressed bonds for a period of days or weeks.

However, unaffected bonds usually remain tradable. Because portfolios are diversified, we typically have positions elsewhere that can be transacted if needed. At fund level, we may apply anti-dilution mechanisms so that entering or exiting investors may bear the transaction costs during stressed conditions.

Where do returns come from over a full cycle?

Over the long term, returns are driven primarily by carry — the coupon income. Spread tightening can contribute over shorter periods, but given the relatively short duration, carry dominates.

Floating-rate income has also been a meaningful contributor when interest rates are elevated. But disciplined loss avoidance is just as important as coupon income.

Where are we in the pricing cycle today?

Spreads widened significantly in late 2022 and early 2023. Since then, in the absence of major industry-changing events, spreads have tightened.

We are not at the peak of the cycle anymore, but nor are we at historic lows. If spreads in certain regions tighten excessively — for example in Japanese earthquake — we may reduce allocation, even if it slightly reduces diversification.

How do you maintain discipline when spreads fall?

We assess pricing adequacy relative to expected loss and volatility. If we believe we are not being adequately compensated, we simply do not invest — regardless of market convention.

We are also active in the secondary market when we see better opportunities there. If, after fees, expected returns approach zero, we reduce exposure. Diversification alone is not sufficient justification for taking uncompensated risk.

When spreads are falling, we do not increase risk to maintain the no-loss return at the same level as before. Instead, we adjust the expected no-loss return downwards to maintain the Expected Loss at fund level.

What differentiates your team?

I believe our long track record is a key strength. We have been active since 2007, through multiple catastrophe cycles. We use proprietary portfolio tools and internal modelling capabilities, in addition to external models. We are also selective — we typically invest in around half of marketed transactions rather than buying the market indiscriminately.

Our team can seem demanding for some structuring banks, as we are scrutinising many factors to fully understand the underlying risks, the modelling but also the cat bonds features such as resets and extensions. If the cat bond is not properly structured from our point of view, we will ask for changes.

Ultimately, I think our differentiation lies in disciplined portfolio construction, structural scrutiny and consistent adherence to pricing adequacy — especially when markets become more competitive.