Catastrophe cover: model layers and reinstatement limits
A catastrophe reinsurance program’s headline size does not tell you how much it will pay after several losses. Attachment points, layer widths, participation, event definitions and available reinstatements determine which loss reaches which protection. Research should reconstruct the sequence before interpreting an insurer’s retained exposure.
The deliverable is a layer schedule and an event-by-event limit ledger. It records covered loss entering the program, gross layer usage, the reinsurer’s share, remaining protection and any separate reinstatement cost. It evaluates disclosed operating protection; it does not decide the legal coverage of a claim or certify reserve adequacy.
Read an actual program as a structured disclosure
Progressive’s 2026–2027 Catastrophe Reinsurance Program exhibit distinguishes the property per-occurrence program, numbered layers, reinstatements and shared aggregate limits. Locate the headings Property per Occurrence Excess of Loss (XOL) Program, Layers 1–3, and Top Layer and Aggregate Shared Limit. The introductory paragraph says the exhibit is an overview rather than a comprehensive description.
That structure supplies the reading discipline, not the numbers used below. A real program can have different regional attachments, multi-year periods, shared limits and other contracts that affect recoveries. Keep the program’s effective dates and covered business beside each layer. Do not apply one property’s protection to all insurance written by the group.
For each layer extract attachment, width, percentage placed, insurer participation, per-occurrence maximum, annual maximum, covered loss basis and restoration terms. Record whether expenses enter the loss amount and whether another protection applies first. A diagram without the supporting definitions may omit a condition material to later-event coverage.
Reproduce a two-event recovery case
Assume a fictional program with all amounts in millions. One layer attaches above 100 of covered loss and has a gross width of 200. Reinsurers cover 90% of the amount reaching the layer; the insurer retains 10% within it as well as losses below and above it. All hypothetical losses are eligible, already measured on the specified basis, and there is no overlapping cover.
For event loss L, gross layer usage before exhaustion is min(max(L − 100, 0), 200). Recovery is 90% of the available covered layer usage. In this example, one full-equivalent reinstatement adds a gross restoration budget of 200 over the contract year. Assume partial restorations are permitted and each unit restored consumes that budget. This is an explicit invented term, not a universal meaning of one reinstatement.
Event 1 causes covered loss of 250. The amount above attachment is 150, below the 200 width. Recovery is 150 × 90% = 135. Retained loss is 250 − 135 = 115, comprising 100 below attachment and 15 participation. Initial gross remaining limit is fifty.
Assume the program restores the 150 used, returning available per-occurrence limit to 200. The unused restoration budget is now fifty. Event 2 causes loss of 400. Although 300 lies above attachment, the layer’s width is only 200. Recovery is 180 and retained loss is 220: 100 below, twenty participation and 100 above the layer.
| Event | Eligible loss | Gross layer used | Recovery at 90% | Retained loss |
|---|---|---|---|---|
| 1 | 250 | 150 | 135 | 115 |
| 2 | 400 | 200 | 180 | 220 |
| Total | 650 | 350 | 315 | 335 |
The identity loss = recovery + retained loss closes in each row and in total. It is not correct to subtract a headline 200 from each event, since the program’s share is 90%, and the first event does not use the full width.
Track what remains for a third event
After Event 2, available gross protection is zero before any further restoration. Under the specifically assumed full-equivalent restoration budget, fifty remains restorable. If restored, it supplies only fifty of gross protection, not another complete 200 layer. An eligible third loss of 250 then receives recovery of 45, retaining 205.
Across the three events, gross usage totals 400 and recoveries total 360. That equals the initial 200 plus 200 restoration budget, multiplied by the 90% share. The annual arithmetic check exposes a common mistake: treating every new event as a fresh full layer despite an exhausted restoration budget.
If the real contract instead permits only one differently defined restoration action, or imposes another aggregate cap, this case is not applicable. Transcribe that actual term and rebuild the sequence. A research model must follow the agreement’s meaning rather than assume the example’s restoration mechanics.
Test a different mechanism: no reinstatement
Keep the first loss of 250, but assume no restoration. After using 150, only fifty remains. Event 2’s potential gross layer amount is 200, yet only fifty is available. Recovery is 45 and retained loss is 355. Two-event recovery is now 180 rather than 315, a difference of 135 entirely attributable to the restoration assumption.
Another contrast is an annual aggregate deductible. Several smaller losses can accumulate toward an annual threshold even if none separately reaches a per-occurrence attachment, depending on the contract. Do not force an aggregate program into the single-event formula. Maintain a cumulative eligible-loss counter with the agreement’s exclusions and cap separately.
These are different protection mechanisms. A company reporting a large annual limit may still retain a severe single event; a per-occurrence program may respond to that event but have less available for the next. Both need a chronological ledger, not a single total-cover ratio.
Separate restoration premiums and recoverability
If a restoration carries an additional premium, record the contractual trigger, formula, amount and any separate protection. For a purely hypothetical fixed premium of five after Event 1, retained loss of 115 does not silently become 110. Show 115 retained claims plus five additional premium, totaling 120 before other items. Do not call the premium another claim recovery.
An expected recovery also differs from cash collected. Add claim notification, estimate, agreement, dispute and payment stages where disclosed. A reinsurer’s obligation to pay depends on the actual contract and covered loss; a modelled recovery is not evidence that the money has already arrived.
Resolve gaps with a contract-specific worksheet
| Field | Obtain | Check |
|---|---|---|
| Loss entering program | Covered-loss definition | Gross/net, expense and regional basis |
| Layer geometry | Schedule and amendments | Attachment, width and participation |
| Prior events | Dated usage history | Remaining limit before current event |
| Reinstatement | Exact restoration terms | Available amount and consumed budget |
| Other programs | Ordering and shared-limit terms | No duplicate recovery |
| Settlement | Recovery and cash records | Estimate separated from collection |
If a filing and investor slide disagree, first check whether one shows placed recovery after participation while the other shows gross layer width. Differences can also reflect renewal dates or Florida-specific versus nationwide coverage. Do not repair them by averaging.
Where loss allocation or treaty terms are absent, report conditional results and the missing clause. Set updates for new events, revised loss estimates, recovery agreements, program renewals and disclosed exhaustion. An insurer can revise a prior event’s estimate and thereby change remaining aggregate protection without a new catastrophe. Preserve the earlier estimate and show the resulting limit change. That is the evidence needed to assess retained exposure without promising a legal or financial outcome.
Preserve the event sequence when estimates change
Create one row per event and one sub-ledger per layer. The row should show opening available limit, potential layer loss, actual gross usage, recovery share, restoration applied and closing remaining limit. A separate annual restoration-budget column prevents the per-occurrence limit from being confused with the total additional capacity purchased for the year.
Sequence matters for attribution. With the same two covered losses of 250 and 400, the no-restoration scenario assigns recoveries of 135 and forty-five. Reverse the order: the 400 loss consumes the entire initial 200 gross limit and receives 180, leaving zero for the later 250 loss. Total recovery remains 180 in this simple case, but the event-level retained results differ. Do not use the annual sum to claim the same event allocation.
If an earlier loss estimate increases, replay the subsequent rows from that point. For example, changing the first loss from 250 to 280 changes its potential gross usage from 150 to 180. Under the main case’s partial-restoration assumption, the restoration needed to refill the layer also rises to 180, leaving only twenty of the annual restoration budget. The second event still uses 200 after restoration, but only twenty could then be restored for a third event. The third loss of 250 would recover eighteen, rather than forty-five.
This is a conditional sensitivity, not a statement that a real contract automatically reallocates settled recoveries. Actual reopening, settlement and restoration provisions control the legal result. Its research purpose is to expose which later-event assumption depends on an earlier estimate. Retain the original estimate, revision date and recalculated balances instead of editing only the annual recovery total.
FAQ
Does a 200-wide layer always reimburse 200?
No. Attachment, participation, eligible loss and available limit all matter.
Does one reinstatement mean unlimited later-event protection?
No. Use the actual restoration amount, definition and any annual cap.