Cat bond pricing explained

Expected loss: the starting point for cat bond pricing

Expected loss (EL) is the probability-weighted average annual loss on a cat bond tranche, expressed as a percentage of notional. It is the single most referenced number in cat bond pricing, and the first thing any investor looks at when evaluating a new deal.

EL comes directly from catastrophe model output. The model simulates tens of thousands of possible years, each containing a set of hypothetical natural catastrophe events. For each simulated year, the model calculates the loss to the bond's layer. EL is the average of those losses across the full simulation set. A bond with an EL of 2% implies that, on average, the investor would lose 2% of principal per year due to covered events.

It is important to understand what EL does not capture. It says nothing about the shape of the loss distribution. Two bonds can have identical ELs but very different loss profiles: one might have a moderate probability of a partial loss, while the other has a tiny probability of a total wipeout. This is why EL alone is never sufficient for pricing, and why investors look at the full loss exceedance curve, not just the mean.

EL also depends entirely on the model vendor, model version, and exposure data fed into it. Different cat model runs on the same bond can produce different ELs. Investors doing their own modelling frequently arrive at a different number than the one in the offering circular, and that disagreement itself influences pricing.

Defining the layer: attachment and exhaustion points

Every cat bond covers a specific layer of loss. The attachment point is the loss threshold at which the bond starts to erode. The exhaustion point is the level at which principal is fully wiped out. The distance between them is the layer width.

Layer position has a direct effect on pricing. A bond that attaches at a very high loss level, say a 1-in-200-year return period, will have a low expected loss because the probability of reaching that threshold is small. But if the layer is narrow, meaning the exhaustion point is close to the attachment point, losses become binary: either the bond is untouched or it is a near-total loss. This cliff risk is something investors price in.

A wider layer that attaches lower absorbs more frequent, smaller losses. The expected loss is higher, the spread is higher, but the loss profile is more graduated. Investors can experience partial losses rather than all-or-nothing outcomes.

Layer position also determines how sensitive the bond is to model uncertainty. Narrow, high-attaching layers sit in the tail of the loss distribution where small changes in model assumptions produce large swings in EL. Lower-attaching, wider layers are more stable because they sit in a part of the curve where the distribution is better constrained by historical data.

When comparing two bonds, looking at EL without understanding where the layer sits in the loss exceedance curve is a mistake. A 1% EL bond attaching at a 1-in-50-year level is a fundamentally different risk from a 1% EL bond attaching at a 1-in-500-year level with a razor-thin layer.

What makes up a cat bond spread?

The spread on a cat bond, the return above the risk-free collateral return, is not a single number derived from a single calculation. It is the sum of several components, each reflecting a different aspect of the risk transfer.

Using risk multiples for relative value

Because cat bonds cover different perils, territories, and layers, investors need a way to compare them on a like-for-like basis. The two most commonly used relative value metrics are the net risk multiple (NRM) and the multiple of expected loss (MEL).

Net risk multiple (NRM)

NRM strips out the expected loss component of the spread and divides the remainder by EL:

NRM = (Spread − EL) / EL

This tells you how much risk compensation you are receiving per unit of expected loss. A bond with a 3% spread and a 1% EL has an NRM of 2.0x. A bond with a 6% spread and a 3% EL also has an NRM of 1.0x. Even though the second bond pays a higher absolute spread, the first bond offers more compensation relative to the risk.

NRM is useful for comparing bonds across different EL levels, but it has limitations. It treats all expected loss as equivalent, which it is not. A 1% EL from US hurricane risk is not the same as a 1% EL from European windstorm because the tail characteristics, model confidence, and correlation profiles are different.

Multiple of expected loss (MEL)

MEL is simpler: MEL = Spread / EL. It measures the total spread relative to expected loss without subtracting EL first. MEL is widely used in the secondary market to compare bonds at different price points and to track how relative value shifts over time. A declining MEL across the market suggests tightening compensation for risk, while a rising MEL indicates investors are demanding more.

From physical probability to risk-adjusted probability

Cat models produce physical probabilities: the objective likelihood of events occurring based on scientific understanding of natural hazards. But investors do not price risk at the actuarial mean. They demand a margin above EL, and the size of that margin depends on the shape of the loss distribution.

The Wang Transform is a mathematical method for converting physical (real-world) loss probabilities into risk-adjusted probabilities. It was developed by Shaun Wang and has become one of the standard approaches for deriving a theoretically justified risk load in cat bond pricing.

In plain terms, the Wang Transform takes the loss exceedance curve and shifts it. Every probability of exceeding a given loss level gets increased by a calibrated amount. Tail events, the large rare losses, get their probabilities increased more than frequent small losses. The result is a new, risk-adjusted exceedance curve that sits above the physical one. The area under this adjusted curve gives you the risk-adjusted expected loss, and the difference between that and the physical EL is the implied risk load.

The transform is controlled by a single parameter, often called lambda. A higher lambda means a larger probability shift and a higher risk load. The market-clearing lambda can be observed by backing it out of actual transaction spreads. When the market is hard (tight capacity, recent losses), the implied lambda rises. When the market softens, it falls.

The Wang Transform matters because it provides a consistent framework for comparing risk loads across different perils and territories. Without it, comparing the risk premium on a Florida hurricane bond to a Japanese earthquake bond becomes subjective. With it, you can at least ask whether the implied probability adjustments are consistent.

Why cat bond spreads change

Cat bond spreads are not static. They move in response to several forces, some driven by the underlying risk and some by market dynamics that have nothing to do with the probability of a hurricane.

Seasonality

Seasonality is the most predictable driver. Spreads on US hurricane-exposed bonds tend to widen as the Atlantic hurricane season approaches (June through November) and tighten after it passes without a major loss. This is rational: during hurricane season, the conditional probability of a loss-causing event is higher. Investors holding these bonds through the season bear more near-term risk and demand compensation for it.

Loss events

A significant catastrophe event reprices the market. Even if a specific bond is not directly affected, a major hurricane or earthquake reminds the market that tail events happen. Post-event, spreads widen across the board as investors reassess risk tolerance and capital availability. Bonds in affected perils and territories see the largest moves.

Supply and demand

New issuance volume matters. Heavy issuance periods force spreads wider because investors have more options and limited capital to deploy. Conversely, when issuance is light, investors competing for a scarce supply of bonds push spreads tighter. The timing of new deals relative to the hurricane season creates predictable patterns in primary market pricing.

Trapped capital

After a loss event, bonds that are exposed to the event but have not yet settled enter a state of uncertainty. Capital backing those bonds is effectively trapped: investors cannot redeploy it, and the eventual recovery is unknown. Trapped capital reduces the effective pool of investable capital in the market, which pushes spreads wider on new and existing bonds.

Reinsurance market conditions

Cat bonds compete with traditional reinsurance and collateralised reinsurance for the same risk. When the broader reinsurance market hardens, pricing for all forms of risk transfer rises, and cat bond spreads follow. When reinsurance capacity is abundant, cat bond spreads compress. The two markets are not perfectly correlated, but they are linked.

How pricing differs between issuance and trading

Cat bond pricing in the primary market (new issuance) and the secondary market (trading of outstanding bonds) follow different dynamics.

Primary market and new issue concession

New cat bonds are priced through a bookbuilding process. The structuring agent sets initial price guidance, investors submit orders, and the final spread is determined by demand. New issues typically include a new issue concession: a premium above where the bond would trade in the secondary market if it already existed. This concession compensates investors for the illiquidity of a new, untested security and for committing capital before the bond has an established trading history.

The size of the new issue concession varies with market conditions. In a soft market with plenty of investor appetite, it can be minimal. In a hard market or for unusual structures, it can be substantial.

Secondary market valuation

Once a bond is issued, its price in the secondary market changes based on updated information. If a new cat model version is released and the bond's EL changes, the price adjusts. If a loss event occurs that affects the bond's peril region, the price adjusts. If the broader market reprices risk, the bond moves with it.

Secondary market valuation involves a tension between mark-to-model and mark-to-market. Mark-to-model values the bond based on its current modelled EL and a prevailing market spread multiple. Mark-to-market values it based on actual bid and offer prices from dealers. For liquid bonds, these should converge. For illiquid bonds or bonds affected by an ongoing event, the two can diverge significantly.

Bid-ask spread

Secondary market cat bonds trade with a bid-ask spread that reflects liquidity, deal size, complexity, and market conditions. Smaller, less well-known issues trade with wider bid-ask spreads. Larger, benchmark-style deals from frequent issuers tend to be tighter. During periods of market stress, bid-ask spreads widen across the board as dealers reduce risk-taking.

Why two cat bonds with the same EL can have different spreads

It is common for two cat bonds with identical expected losses to trade at different spreads. This is not an anomaly; it reflects genuine differences in risk characteristics that EL alone does not capture.

How ILS101 teaches pricing

ILS101 covers pricing through the full chain: from catastrophe model output, through expected loss calculation, spread decomposition, risk multiples, and secondary market valuation. The course explains not just the mechanics but the reasoning behind why the market prices risk the way it does, and why model output alone is never the full story.

• Expected loss calculation from cat model output
• Spread decomposition: EL, risk load, and expense load
• Net risk multiple and MEL for relative value comparison
• The Wang Transform and risk-adjusted pricing
• Primary issuance mechanics and new issue concession
• Secondary market valuation: mark-to-model vs mark-to-market
• Why trigger type, peril, territory, and tail risk create pricing differences
• Supply, demand, trapped capital, and reinsurance market influence on spreads