Attachment Point
Definition
Attachment point is the loss threshold at which an insurance-linked security, reinsurance layer, catastrophe bond, or structured risk position begins to absorb losses.
Why it matters
Attachment point determines how remote or exposed a risk layer is. A low attachment point may pay a higher spread because losses reach it more often. A high attachment point may look safer but still carry severe tail risk if the modeled event occurs. Investors need to understand where their layer sits in the loss stack, not only the headline coupon.
Common misconceptions
- •A high attachment point does not mean no risk; it means the position is exposed only after losses exceed a threshold.
- •Attachment point is not the same as expected loss; expected loss depends on event probability, severity distribution, exhaustion point, and trigger design.
- •Modeled attachment can differ from realized attachment if definitions, covered perils, reporting, or industry-loss estimates move after an event.
- •A wider spread is not necessarily attractive if the layer attaches frequently or if model uncertainty is high.
Technical details
Position in the loss stack
Insurance and reinsurance risk is often divided into layers. The cedent retains the first portion of losses, then one or more risk-transfer layers absorb losses above defined thresholds. The attachment point is the threshold where a given layer starts paying. The exhaustion point is where the layer is fully used. A layer attaching at $500 million and exhausting at $700 million has $200 million of limit exposed after the first $500 million of covered loss.
Occurrence versus aggregate attachment
Occurrence layers attach based on a single covered event, such as one hurricane or earthquake. Aggregate layers attach after cumulative losses across a period exceed a threshold. Aggregate structures can be exposed to multiple medium events even if no single event is catastrophic. Investors should distinguish event severity risk from frequency risk because the same attachment point can behave differently under occurrence and aggregate definitions.
Expected loss connection
Expected loss is the probability-weighted average loss to the layer. Attachment point influences expected loss, but it does not determine it alone. A low layer in a low-risk region can have lower expected loss than a high layer in a highly exposed region. Model output should show expected loss, attachment probability, exhaustion probability, and sensitivity to event frequency and severity.
Return period language
Attachment is often described with return-period language, such as a one-in-100-year or one-in-250-year event. That language can be misleading because it does not mean the event can happen only once per century. It means the modeled annual probability is roughly 1% or 0.4%, subject to model assumptions. Investors should translate return periods into annual probability, cumulative multi-year probability, and stress scenarios.
Trigger design
The attachment point can be tied to indemnity losses, industry-loss indexes, parametric measurements, or modeled loss formulas. Indemnity triggers follow the sponsor's actual losses but can involve reporting lag and moral hazard. Industry-loss triggers reduce sponsor-specific loss adjustment but introduce basis risk. Parametric triggers can settle faster but may pay differently from actual economic loss.
Spread and risk premium
Investors compare spread to expected loss to estimate risk premium. A bond with 8% spread and 2% expected loss has a 4.0x multiple of expected loss. That multiple is not automatically sufficient. It should be judged against model uncertainty, peril type, seasonality, collateral yield, liquidity, secondary-market depth, and the chance of trapped collateral after an event.
Seasonality and time at risk
Some perils are seasonal. A hurricane-exposed layer issued just before peak season carries different near-term risk than the same annualized spread after the season has passed. Secondary-market pricing often reflects time at risk, event forecasts, and recent catastrophe activity. Attachment analysis should be paired with the remaining risk period, not only annual model output.
Model uncertainty
Catastrophe models estimate event frequency, severity, vulnerability, and financial terms. Small changes in assumptions can shift the probability that losses reach the attachment point. Climate trends, insured-value inflation, litigation, demand surge, and construction-cost inflation can all make historical loss data less reliable. Attachment should therefore be analyzed across model versions and stress scenarios.
Exhaustion point and layer thickness
A layer's risk is shaped by both attachment and exhaustion. Thin layers can be fully exhausted by a narrow band of losses, while thick layers absorb losses over a broader range. A thin high-attaching layer may have low probability but binary outcomes. A thicker lower layer may experience partial losses more often. Recovery behavior, mark-to-market volatility, and diversification differ across these designs.
Collateral and trapped capital
Cat bonds and collateralized reinsurance often hold collateral to support loss payments. After a potential covered event, collateral can be trapped while losses develop. The layer may not be declared impaired immediately, but investors can lose liquidity and reinvestment flexibility. Attachment analysis should include event reporting timelines and conditions for releasing collateral.
Basis risk
The layer can attach differently from an investor's intuitive view of the event because triggers define losses in a specific way. An industry-loss trigger may attach even when the sponsor's own losses are lower, or fail to attach when the sponsor suffers losses but the industry estimate is below threshold. Parametric structures can behave similarly if wind speed, earthquake intensity, or location measurements do not match actual insured damage.
Loss creep and development
A layer may appear safe immediately after an event and then become impaired as claims develop. Construction inflation, litigation, social inflation, business-interruption claims, and late reporting can push losses upward. Investors should understand how the attachment point is tested over time, how collateral is trapped during uncertainty, and whether modeled losses include realistic development assumptions.
Portfolio construction
A portfolio can appear diversified by issuer while concentrating in similar attachment bands, peril zones, or event seasons. Investors should review aggregate exposure by peril, geography, sponsor, trigger type, return period, and attachment probability. Multiple positions attaching around the same industry-loss level can create correlation that is invisible from coupon and issuer lists alone.
Diligence checklist
Review attachment and exhaustion points, covered perils, trigger definition, expected loss, attachment probability, exhaustion probability, return period, model vendor output, sensitivity cases, collateral terms, event reporting process, and historical loss experience for comparable layers. The goal is to understand when principal starts taking losses and how quickly partial impairment can become full impairment.
Practical investor interpretation
Attachment point should be translated into plain-language event scenarios. For example: what type of hurricane, earthquake, wildfire season, mortality event, or industry loss would be needed to reach this layer? If the answer is unclear, the structure is not yet diligence-ready. The best analysis connects abstract thresholds to realistic stress events and sponsor-specific exposure.
Monitoring after issuance
Attachment analysis does not stop at purchase. Investors should monitor insured-value growth, sponsor exposure changes, model updates, climate signals, loss creep, collateral releases, and secondary marks after major events. A layer that looked remote at issuance can become more exposed if insured values grow, inflation raises repair costs, or updated models push more loss into the covered band.