Marine War Risk Insurance in 2026: Hormuz, Red Sea and the New Underwriting Reality

The 2026 capacity position

On 19 June 2026 Lloyd's announced a Chubb-led consortium making available up to USD 200 million of capacity separately for hull and protection and indemnity risks, together with a dedicated USD 200 million of cargo capacity. The capacity is offered subject to underwriting criteria, sanctions requirements and applicable regulation, and it is written through the Lloyd's market rather than as a bilateral facility.

For an underwriter the announcement matters less as a headline number than as a change in the structure of available cover. Separating hull and P&I capacity from cargo capacity recognises that these are different risks with different claimants, different duration and different reinsurance behaviour. It also allows a shipowner and a cargo interest to obtain cover without competing for the same limit.

• Line sizes: larger consortium capacity permits smaller individual signings for the same programme, which reduces single-risk concentration.
• Attachment: reinsurance and facultative structures can be rebuilt around a higher and more predictable primary layer.
• Continuity: consortium capacity that persists through a period of tension is worth more to an assured than opportunistic capacity that withdraws.
• Discipline: capacity offered subject to underwriting criteria and sanctions screening is not the same as capacity offered at any price.

Hormuz and the Red Sea are not the same risk

It is tempting to treat both corridors as a single geopolitical exposure. The underwriting characteristics differ enough that a common treatment produces the wrong price and the wrong accumulation figure.

The practical consequence is that time in area, not transit count, should drive frequency in the Red Sea, while blocking-and-trapping and detention wordings drive the tail in Hormuz. A single blended frequency parameter fitted across both corridors will understate the Hormuz tail and misprice Red Sea transits with long exposure windows.

Hull, cargo and P&I: three different exposures

War risk is written across three principal covers that share a peril but not a loss profile. Aggregating them on one limit basis, as some portfolio reports still do, hides the real capital consumption.

• Hull war: severity bounded by the insured value, with total loss, constructive total loss and detention as the main outcomes. Exposure is per vessel and per transit.
• Cargo war: many small to medium exposures moving simultaneously, often across multiple vessels and multiple assureds, with severity driven by commodity value and delay consequences rather than by hull value.
• P&I war: third-party liability, crew injury and death, wreck removal, pollution and passenger liabilities, typically excluded from mutual cover above defined limits and placed in the commercial war market.

The three respond differently to the same event. A missile strike on a laden tanker can produce a hull total loss, a cargo total loss with pollution consequences, and a P&I claim for crew injury and wreck removal simultaneously. An accumulation model that does not link the three by event will report three unrelated exposures and one badly understated single-event loss.

This is also why the separation of consortium capacity into hull and P&I on one side and cargo on the other is analytically sensible. The two groups have different claim tails, different claimants and different reinsurance recovery patterns.

Technical pricing under sparse data

The pricing chain for a war-risk transit remains the same in 2026 as it has always been. What has changed is the weight that has to be placed on judgement and on the uncertainty loading, because the regime shifts faster than any credible dataset accumulates.

P = E[L] + kappa * sigma_L + C * r_c + Expenses

Two components deserve emphasis. First, frequency should be expressed per vessel-hour or vessel-day in a defined zone rather than per transit, so that a twelve-hour passage and a three-day port call inside the zone are not treated as equivalent. Second, the capital term is marginal, not average: the same transit is worth materially more capital when the corridor already carries a large aggregate that day.

1. Define the zone and the exposure window precisely in the wording, including port calls and waiting time.
2. Set frequency from a documented view combining incident history, current conditions and expert judgement, with the judgement recorded separately from the data.
3. Model severity separately by loss type: total loss, partial damage, detention and consequential delay.
4. Apply the uncertainty loading explicitly, and disclose it to the underwriter rather than burying it in the frequency assumption.
5. Charge the marginal capital cost given the corridor aggregate on the intended sailing date.

Exposure aggregation and the measurement basis

Accumulation control is where war-risk portfolios succeed or fail. Individually well-priced transits combine into an unsound portfolio when many of them sit in the same corridor on the same day.

• Measure corridor aggregate on a rolling window, at least daily, not on month-end snapshots that miss scheduling clusters.
• Aggregate by event, linking hull, cargo and P&I exposures on the same vessel and the same voyage.
• Hold zone limits and rolling-window limits with underwriting authority tied to remaining headroom.
• Track exposure after binding, since routing decisions frequently change once a voyage is under way.
• Reconcile the modelled aggregate to the reinsurance programme structure, so that the attachment point is expressed on the same basis as the measurement.

Additional consortium capacity does not remove this problem. It can worsen it, because a market with more available limit tends to write more of the same exposure at larger aggregate before the accumulation control binds. Capacity should therefore be accompanied by a tightening of the measurement basis, not a relaxation of it.

Capital, ORSA and stress testing

A marine war-risk book belongs in the ORSA as a named scenario rather than as a line item inside a general non-life stress. The scenario has to be specific enough that the answer is a number.

1. Define the event: corridor, duration, number of vessels affected and the loss types triggered.
2. Apply it to the live book on a realistic peak date, not the average exposure.
3. Compute the gross loss by class, then apply the actual reinsurance wordings, including any war exclusions or event definitions that may be contested.
4. State the net loss, the effect on own funds and the movement in the solvency ratio.
5. Identify management actions with the time required to execute them, and confirm they are credible under the same stressed conditions.

Reverse stress testing is particularly informative here. Asking what corridor event would render the marine account unviable usually reveals that the binding constraint is not severity per vessel but the number of simultaneous exposures the accumulation controls permit.

Model governance for a judgement-heavy class

War-risk pricing and accumulation models should be tiered at the top of the model inventory. They are material, they are judgement-heavy, and the environment they describe is not stationary, which means backtesting provides limited comfort.

• Record the model owner, the validator and the date of the last independent review for every war-risk model in the inventory.
• Document every expert-judgement parameter with its rationale, its source and the date it was last challenged.
• Log overrides of model output with the reason, the approver and the outcome, and review the override log for systematic bias.
• Validate the accumulation engine separately from the pricing model; they fail in different ways.
• Re-review parameters on a defined event trigger, not only on the annual cycle, since a corridor can change character within days.
• Keep sanctions and regulatory screening as a hard control in the underwriting workflow, not a post-bind check.

The independent validation standard applied to credit and market risk models transfers directly. The difference is that in war risk the validator cannot rely on statistical performance testing, so the review must concentrate on conceptual soundness, the defensibility of judgement, data lineage and the behaviour of the model under scenarios rather than on fit statistics.