Large Loss, Catastrophe, and Reinsurance Adjustments in Filings

Why Loss Adjustments Need a Consistent Basis

Historical losses may include an unusually severe individual claim, a catastrophe, or recoveries under a reinsurance program. Those amounts are part of the observed experience, but they may not represent the losses expected under the proposed rates and prospective coverage. A filing adjustment should therefore explain how historical experience is made comparable to the future period—not simply remove an inconvenient result.

The adjustment depends on the question being answered. A large individual loss might be capped at a selected threshold and replaced by an expected provision for losses above that threshold. Catastrophe losses might be replaced by a selected long-term provision. Reinsurance recoveries might be added back to establish a gross basis, then prospective recoveries deducted to estimate future retained losses. The selected method should match the exposure base and loss basis used elsewhere in the indication.

Definition: An adjusted historical loss is an observed loss amount modified by a documented method to make it more representative of the prospective coverage, loss level, and reinsurance basis. An adjustment should identify the historical component changed, the replacement amount or method, the evidence supporting the selection, and the effect on the proposed rate level.

An adjustment is not the same as deleting a loss. If large claims are removed above a threshold, the resulting experience no longer includes the expected cost of those claims unless a provision for them is added. Similarly, catastrophe losses are a component of insured risk; eliminating observed catastrophe losses without a prospective catastrophe provision can understate expected cost. Adjustments also should not duplicate loss development, trend, or other changes already made to the experience.

Replacing Large-Loss Experience

A common approach is to identify the portion of each historical loss above a selected threshold, remove that portion from the historical total, and add a selected provision for the expected cost of that layer. The threshold and provision may be selected using the purpose of the analysis, credible claim data, exposure volume, and other relevant actuarial evidence. A cap is a tool for stabilizing experience; it is not itself an estimate of the prospective cost of losses above the cap.

L_{\text{adjusted}} = \sum_{j=1}^{n}\left(L_j-E_j\right) + \sum_{j=1}^{n}P_j

Here $L_j$ is the historical loss for period $j$, $E_j$ is the identified amount being replaced, and $P_j$ is the selected provision for that period. If the method caps individual claims at threshold $c$, then for a claim amount $X$, the amount removed is $\max(X-c,0)$. The provision must represent the expected cost of the layer removed, on a basis consistent with the rest of the losses. If the selected provision is stated per exposure, multiply it by the applicable number of exposures before adding it to the loss total.

Identify the component.
Specify which claims or claim portions are treated as large losses, and define the threshold or other selection rule.

Remove only that component.
Subtract the identified amount from historical losses. Do not subtract the entire claim if only its excess layer is being adjusted.

Add the selected provision.
Use a supported prospective estimate for the component removed, applying the correct exposure count if the provision is per exposure.

Check the basis and effect.
Confirm that loss development, trend, exposure, and reinsurance treatments are consistent, then compare adjusted and unadjusted loss costs.

Worked Example: Replacing Large-Loss Amounts With an Annual Provision

A hypothetical liability portfolio has three years of losses of $300,000, $360,000, and $270,000. Identified large-loss components within those totals are $100,000, $180,000, and $60,000, respectively. The selected provision replacing those components is $50,000 per year. Assume all amounts are already at the same development and trend level, and no further adjustment is needed.

Method. Subtract each identified large-loss component from its year, then add the selected annual provision of $50,000 to each year.

Year 1 remaining losses are $300,000 − $100,000 = $200,000. After the provision, adjusted losses are $200,000 + $50,000 = $250,000.

Year 2 remaining losses are $360,000 − $180,000 = $180,000. After the provision, adjusted losses are $180,000 + $50,000 = $230,000.

Year 3 remaining losses are $270,000 − $60,000 = $210,000. After the provision, adjusted losses are $210,000 + $50,000 = $260,000.

The adjusted total is $250,000 + $230,000 + $260,000 = $740,000. As a check, unadjusted losses total $300,000 + $360,000 + $270,000 = $930,000. The removed components total $100,000 + $180,000 + $60,000 = $340,000, and the three provisions total $150,000. Therefore, $930,000 − $340,000 + $150,000 = $740,000.

Suppose the three years contain 1,000, 1,200, and 900 exposures. The total is 3,100 exposures, so the adjusted loss cost is $740,000 ÷ 3,100 = $238.7097 per exposure, or $238.71 to the nearest cent. The unadjusted loss cost is $930,000 ÷ 3,100 = $300.00 per exposure. The adjustment reduces the historical loss cost by $61.29 per exposure. The common mistake avoided is subtracting the large-loss components without restoring any expected cost for them.

Normalizing Catastrophe Experience

Catastrophes can create substantial variation across short historical periods. A filing may use a selected catastrophe provision instead of treating the observed catastrophe amount as a reliable estimate of the prospective period. The selection should be supported by appropriate information, such as relevant historical experience, exposure characteristics, catastrophe modeling, or other suitable evidence. The method should make clear whether the provision is an annual amount, a rate per exposure, or another measure.

A catastrophe adjustment is not automatically appropriate merely because a year had a large event. The actuary should consider whether the historical experience and prospective risk are comparable, whether catastrophe costs are already included elsewhere, and whether the selected provision applies to the same coverage and exposure base. A short experience period with no catastrophe loss does not, by itself, establish that the prospective catastrophe cost is zero.

Worked Example: Replacing Observed Catastrophe Losses

A hypothetical property program has three years of losses of $420,000, $310,000, and $500,000. Identified catastrophe amounts are $240,000, $0, and $320,000. A supported selection is an annual catastrophe provision of $90,000. Each year has 1,000 exposures. Assume all losses are on a common development and trend basis, and the selected provision applies to these exposures.

Method. Remove each observed catastrophe amount and add the $90,000 annual provision.

Year 1 non-catastrophe losses are $420,000 − $240,000 = $180,000. Adjusted losses are $180,000 + $90,000 = $270,000.

Year 2 non-catastrophe losses are $310,000 − $0 = $310,000. Adjusted losses are $310,000 + $90,000 = $400,000.

Year 3 non-catastrophe losses are $500,000 − $320,000 = $180,000. Adjusted losses are $180,000 + $90,000 = $270,000.

The adjusted total is $270,000 + $400,000 + $270,000 = $940,000. The check is $1,230,000 of observed losses − $560,000 of identified catastrophe losses + $270,000 of provisions = $940,000. Across 3,000 exposures, the adjusted loss cost is $940,000 ÷ 3,000 = $313.33 per exposure to the nearest cent. The unadjusted loss cost is $1,230,000 ÷ 3,000 = $410.00 per exposure. The adjustment lowers the loss cost by $96.67 per exposure in this example, but it does not imply that catastrophe risk has been removed: the selected provision remains in the adjusted losses.

Aligning Historical and Prospective Reinsurance

Losses may be reported gross of reinsurance or net of reinsurance recoveries. Gross losses include the covered loss amounts before recoveries; net losses subtract the recoveries attributed to those losses. A filing analysis must state which basis it uses and apply it consistently. If historical losses are net but the prospective reinsurance arrangement differs, historical net losses may not represent the expected future retained losses.

One clear method is to begin with gross losses, project them on the applicable basis, and subtract recoveries expected under the prospective reinsurance program. Another is to adjust historical net losses to reflect the difference between historical and prospective reinsurance. These are alternative routes to a consistent estimate; do not both add back historical recoveries and then subtract the same recoveries a second time. Reinsurance premiums and related costs are distinct from claim recoveries and should not be silently included in the loss adjustment.

\text{Prospective net losses} = \text{prospective gross losses} - \text{prospective recoveries}

This identity follows from the definition of net losses: for a given set of covered claims, the insurer's retained claim cost equals the gross claim cost less the amount recovered from reinsurers. The identity applies only when gross losses and recoveries refer to the same claims, coverage, period, and valuation basis. If a recovery is unavailable, disputed, or subject to a different limit or attachment, the analysis must reflect that circumstance rather than treating it as an unconditional offset.

Worked Example: Converting Historical Gross Losses to a Prospective Net Basis

A hypothetical program has 3,000 exposures. Its historical gross losses are $600,000, $700,000, and $500,000, for a total of $1,800,000. Recoveries under the historical reinsurance program are $0, $150,000, and $50,000, for a total of $200,000. The prospective program is expected to produce $75,000 of recoveries on an equivalent exposure basis. Assume no additional trend, development, or exposure adjustment is required.

Method. First establish the gross historical loss cost, then apply the prospective recovery estimate to obtain the prospective net loss cost.

Historical net losses are $600,000 − $0 = $600,000 in Year 1, $700,000 − $150,000 = $550,000 in Year 2, and $500,000 − $50,000 = $450,000 in Year 3. They total $1,600,000. The gross total is independently checked as $600,000 + $700,000 + $500,000 = $1,800,000.

The gross loss cost is $1,800,000 ÷ 3,000 = $600.00 per exposure. Prospective net losses are $1,800,000 − $75,000 = $1,725,000, giving a prospective net loss cost of $1,725,000 ÷ 3,000 = $575.00 per exposure.

For comparison, simply using historical net losses would give $1,600,000 ÷ 3,000 = $533.33 per exposure. That amount reflects the historical recovery pattern, not the specified prospective recovery estimate. The common mistake avoided is subtracting the prospective $75,000 from historical net losses; doing so would subtract recoveries once in the historical net amount and again in the prospective adjustment. A filing should separately identify any reinsurance cost not included in claim recoveries.

Why the Adjustment Formula Works—and What to Document

The large-loss formula can be justified by separating observed losses into two pieces: the portion retained in the historical experience and the portion replaced. For each period, write the observed total as $L_j=(L_j-E_j)+E_j$. The adjusted estimate keeps the first term and substitutes $P_j$ for the second. Thus, $L_j-E_j+P_j$ is the retained historical amount plus the selected estimate of the removed component. Summing over periods gives the adjustment formula. This reasoning also shows why subtracting $E_j$ alone is not a neutralization: it assigns zero cost to the replaced layer.

The amount of an adjustment is important, but a reviewer also needs to understand why it was selected and how it changes the result. Filing support should make the calculation reproducible and connect it to the proposed rate level. Useful documentation includes:

The experience period, loss valuation date, exposure base, and whether losses are gross or net of reinsurance.
The large-loss threshold or catastrophe identification rule, with a claim-level or event-level reconciliation to recorded totals.
The source, method, and assumptions supporting the replacement provision, including whether it is per year or per exposure.
The treatment of development, trend, recoveries, and any amounts included elsewhere, to show that adjustments are neither omitted nor duplicated.
The unadjusted and adjusted totals, the corresponding loss costs, and the resulting change in the indication or proposed rate level.
Any sensitivity that is material to the selection, especially where a threshold or catastrophe provision has a meaningful effect on the result.

For example, if the adjusted loss cost changes by $a$ dollars per exposure and the relevant premium or rate calculation uses the same exposure basis, the filing should show how that difference enters the loss component of the indication. It should not describe the loss-cost change as the total rate change unless the other components and the indication formula support that conclusion. The objective is a traceable adjustment, not an unsupported claim that one method is uniquely correct.

Common Exam and Filing Traps

Deleting the entire claim. If the method removes only the amount above a threshold, retain the portion below it.
Removing losses without replacing expected cost. A large-loss or catastrophe provision is needed when the selected method is intended to represent the expected cost of that component.
Mixing gross and net bases. Do not compare gross historical losses with prospective net losses without an explicit reconciliation.
Double-counting recoveries. Start from a clearly stated gross or net basis and apply the prospective reinsurance adjustment once.
Using the wrong unit. A provision per exposure must be multiplied by the matching exposure count; an annual dollar provision should not be multiplied by exposures again.
Ignoring other adjustments. Check whether the same loss component has already been addressed through development, trend, or another provision.

Key takeaway: Adjust historical large losses, catastrophes, and reinsurance only through a stated, supported method that puts the experience on a prospective and internally consistent basis. Show the arithmetic, evidence, and effect on the loss cost or indication, and verify that no component is omitted or counted twice.

Check Your Understanding

For each problem, state the adjustment method and check the final total or rate basis.

Three years have losses of $250,000, $330,000, and $280,000. Identified large-loss components are $70,000, $120,000, and $40,000. Replace those components with a selected provision of $35,000 per year. What is the adjusted three-year total? (Answer: $735,000)
Two years of hypothetical property losses are $500,000 and $400,000, including catastrophe amounts of $300,000 and $0. Replace the catastrophe amounts with an $80,000 annual provision. What is the adjusted total? (Answer: $760,000)
Historical gross losses are $700,000 over 1,000 exposures. Expected prospective recoveries on the same exposure basis are $40,000. What is the prospective net loss cost per exposure, assuming no other adjustments? (Answer: $660.00 per exposure)
Three claims are $90,000, $160,000, and $220,000. A method caps each claim at $100,000 and replaces the total excess with a selected provision of $45,000 per year for three years. What is the adjusted total? (Answer: $425,000)
A filing starts with $900,000 of historical gross losses and $120,000 of historical recoveries. Prospective recoveries are selected at $50,000 on the same basis. What is the prospective net loss estimate if gross losses require no other adjustment, and what double-counting error should be avoided? (Answer: $850,000; do not subtract prospective recoveries from historical net losses after already subtracting historical recoveries.)

What You'll Learn