Portfolio Steering by Peril, Region, and Occupancy

Introduction

Catastrophe portfolio steering is the process of shaping what the insurer writes, renews, or reduces so that the resulting portfolio remains within its risk appetite. A portfolio can be within a total exposure limit yet still be concentrated in one peril, region, or occupancy. Conversely, a segment with a high modeled loss rate may represent only a small part of the total portfolio. A useful decision therefore considers both how much exposure is present and what modeled losses that exposure can produce.

Steering is not simply a ranking of segments from “safe” to “risky.” It requires consistent exposure definitions, appropriate modeled measures, and constraints that apply to the decision being made. The examples below use hypothetical portfolios and modeled results. They are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Definition: A portfolio segment is a specified part of the insured portfolio, such as a peril-region-occupancy combination. Exposure may be measured by total insured value, policy count, or another relevant quantity, but comparisons require a consistent measure and time basis. A modeled loss measure is a result produced by applying a catastrophe model to a defined portfolio and set of assumptions. Examples include annual average loss (AAL) and a specified return-period portfolio loss.

Choose Measures That Match the Decision

An exposure total describes the amount insured, not the full risk of that amount. Two regions with the same total insured value can have different modeled losses because of differences in hazard, vulnerability, coverage, and exposure characteristics. For this reason, steering commonly reviews several measures together.

Exposure amount and share: the amount in a segment and its fraction of the corresponding portfolio total.
AAL and modeled loss rate: AAL summarizes modeled average annual loss under the model assumptions. Dividing segment AAL by segment exposure gives a rate useful for comparing segments when the exposure measure and model basis are consistent.
Specified event or return-period loss: a loss measure used to test whether the portfolio remains within a defined accumulation or risk limit.
Loss contribution and concentration: the share of a modeled loss measure arising from a segment, interpreted with care if the measure is a quantile of an aggregate distribution.

A modeled loss rate is a comparison tool, not a complete underwriting decision. It does not by itself capture all relevant constraints, and it should not be compared across segments if the exposure denominators, coverage basis, or model assumptions differ. Likewise, a portfolio limit may apply to a modeled portfolio loss rather than to the sum of segment losses at the same return period.

Key result: If segments form a disjoint partition of the insured losses and their expected losses are measured on the same basis, portfolio AAL is the sum of segment AALs. This follows from linearity of expectation: for segment losses $X_1,\ldots,X_n$, $E[\sum_j X_j]=\sum_j E[X_j]$. Independence is not required. In contrast, a portfolio quantile or return-period loss is generally not the sum of segment quantiles, because the segments’ losses may occur together or separately.

For additive AAL, an exposure-weighted portfolio loss rate follows directly. Let $E_j$ be the exposure in segment $j$, and let $r_j=E[X_j]/E_j$ be its modeled AAL per unit of exposure. When the segment exposures sum to the portfolio exposure, then

r_{\text{portfolio}} =\frac{\sum_j E[X_j]}{\sum_j E_j} =\frac{\sum_j E_j r_j}{\sum_j E_j}.

The portfolio rate is therefore an exposure-weighted average of segment rates, not an unweighted average. The identity relies on consistent segment definitions and exposure units; it is not a formula for combining return-period losses.

Turn Risk Appetite Into Steering Constraints

A practical steering decision starts by identifying the dimension of the constraint. A limit may apply to a peril-region event loss, an occupancy share, a location or zone exposure total, or a modeled loss for the whole portfolio. The proposed change must be evaluated against each applicable constraint. Passing one test does not imply that the change passes the others.

Define the proposed change.
Specify the peril, region, occupancy, exposure amount, coverage basis, and whether the decision is new business, renewal, or reduction.

Identify the binding measures.
List relevant exposure limits, concentration limits, and modeled loss constraints. Confirm that each measure uses the intended model version and portfolio scope.

Evaluate the post-change portfolio.
Use an appropriate model rerun when the measure depends on the whole portfolio. Use a segment calculation only when the constraint and its assumptions support that calculation.

Apply all constraints and document the decision.
Approve, limit, reshape, or decline the change according to the tightest applicable constraint, and record the assumptions and remaining headroom.

A simple capacity calculation is valid when the incremental modeled loss is assumed to scale linearly with added exposure and the existing portfolio result is held fixed. If $L_0$ is the current loss measure, $c$ is the modeled loss per unit of added exposure, $x$ is the proposed exposure, and $K$ is the applicable limit, the condition is $L_0+cx\leq K$. Solving gives the maximum permissible addition $x\leq(K-L_0)/c$, provided $c>0$ and the stated linearity assumption is reasonable. If the model indicates material interactions or the limit is a portfolio quantile, a portfolio rerun is preferable to treating this calculation as exact.

Worked Example: Limiting Growth in a Wind-Exposed Region

A hypothetical insurer has $120 million of insured value in a northern region exposed to wind. Its modeled AAL for that segment is $0.72 million, and its modeled 1-in-100 loss is $24 million. The region’s 1-in-100 loss limit is $28 million. A proposed account adds up to $30 million of insured value. Under the stated screening assumption, the account adds $0.008 of AAL and $0.20 of 1-in-100 loss per dollar of insured value. Determine the maximum addition allowed by the event-loss limit and calculate the resulting exposure and AAL.

Method. First calculate the event-loss headroom, then divide it by the assumed incremental event-loss factor. The factor is $0.20 of modeled event loss per $1 of exposure.

The available headroom is $28-24=4$ million dollars. Thus the maximum addition under the event-loss limit is

x_{\max}=\frac{4}{0.20}=20\text{ million dollars}.

The full $30 million request would add $30(0.20)=6$ million dollars of modeled event loss, giving $24+6=30$ million dollars, which exceeds the $28 million limit. At the permitted $20 million, the added event loss is $20(0.20)=4$ million dollars, so the resulting modeled event loss is $24+4=28$ million dollars.

The resulting insured value is $120+20=140$ million dollars. Added AAL is $20(0.008)=0.16$ million dollars, so resulting segment AAL is $0.72+0.16=0.88$ million dollars. The calculations check: event loss is exactly at the limit, and the maximum addition is below the $30 million request. This screening result depends on the given linear scaling assumption; it is not a substitute for a full rerun if that assumption does not hold.

Steer Across Occupancies as Well as Geography

Peril and region do not tell the whole accumulation story. Occupancy can affect both the amount and vulnerability of exposed property. A portfolio may therefore set an occupancy share limit in addition to a modeled event-loss limit. When the proposed business belongs to a constrained occupancy, the share denominator can also change as the business is added. The post-change share—not just the added amount—must be tested.

Worked Example: Applying an Occupancy Share and an Event-Loss Limit

A hypothetical coastal portfolio has $100 million of insured value: $50 million residential, $30 million industrial, and $20 million warehouse. Its modeled loss under a specified cyclone event is $16 million, made up of $5 million residential, $6 million industrial, and $5 million warehouse loss. The occupancy limit requires industrial insured value to be no more than 35% of total portfolio insured value. The event-loss limit is $20 million. A proposed industrial account has insured value $x$, and the screening model assumes it adds $0.25 of cyclone-event loss per dollar insured. Find the maximum $x$ satisfying both limits.

Occupancy test. After adding $x$, industrial exposure is $30+x$ million dollars and total exposure is $100+x$ million dollars. The share constraint is

\frac{30+x}{100+x}\leq 0.35.

Multiplying by the positive denominator $100+x$ gives $30+x\leq35+0.35x$. Therefore, $0.65x\leq5$, or $x\leq5/0.65=7.6923$ million dollars.

Event-loss test. The event-loss headroom is $20-16=4$ million dollars. At $0.25 of event loss per dollar of added insured value, the event constraint gives $0.25x\leq4$, so $x\leq16$ million dollars.

Both conditions must hold, so the occupancy limit binds: the mathematical maximum is $7.6923 million. Rounded to two decimals, the maximum is $7.69 million. At the unrounded maximum, total exposure is $107.6923$ million dollars and industrial exposure is $37.6923$ million dollars, giving a share of exactly 35%. Added event loss is $0.25(7.6923)=1.9231$ million dollars, so total modeled event loss is $16+1.9231=17.9231$ million dollars, below the $20 million limit. The decision is checked against both constraints; passing only the event-loss test would incorrectly permit too much industrial exposure.

Use Portfolio-Level Results for Renewal Choices

For a renewal or a set of coordinated portfolio changes, segment summaries help explain where risk sits, but a portfolio-level model result is needed when the governing limit applies to portfolio loss. This is particularly important for return-period losses: summing a region’s and an occupancy’s modeled 1-in-100 losses can double-count overlapping exposure, while adding separate segment quantiles can misrepresent dependence. The model’s portfolio result should be calculated on the post-change portfolio using a consistent basis.

Worked Example: Selecting a Renewal Mix Under Multiple Limits

A hypothetical insurer compares three alternative renewal mixes. Each row below is a separate modeled rerun of the complete post-renewal portfolio. The insurer’s constraints are portfolio 1-in-100 loss no greater than $40 million, coastal-region 1-in-100 loss no greater than $24 million, and wood-frame share of total insured value no greater than 30%. AAL is reported for context but is not one of these three limits.

Alternative	Portfolio AAL	Portfolio 1-in-100 loss	Coastal 1-in-100 loss	Wood-frame exposure share
A	$1.50 million	$39 million	$23 million	28%
B	$1.42 million	$41 million	$21 million	29%
C	$1.62 million	$38 million	$25 million	28%

Method. Test every alternative against every binding limit. Alternative A has $39\leq40$, $23\leq24$, and $28\%\leq30\%$, so it satisfies all three. Alternative B has a portfolio loss of $41 million, exceeding its $40 million limit, even though its coastal loss and wood-frame share pass. Alternative C satisfies the portfolio loss and wood-frame constraints, but its $25 million coastal loss exceeds the $24 million limit.

On the stated constraints, A is the only feasible alternative. B’s lower AAL does not make it acceptable under the portfolio loss limit; C’s lower portfolio loss does not cure its coastal-limit breach. This comparison uses the complete modeled portfolio outputs rather than adding segment quantiles. The results check directly against each stated threshold.

Common Steering Traps

Using exposure alone as a risk proxy. Equal insured values need not have equal modeled losses. Use exposure to track accumulation, and modeled loss measures to assess modeled risk.
Taking a simple average of segment loss rates. The portfolio rate is exposure-weighted when segment AALs and exposures share a consistent basis.
Adding segment return-period losses as if they were portfolio losses. Segment quantiles generally do not add to the portfolio quantile. Use a portfolio model result when the constraint is defined at portfolio level.
Checking only the most visible constraint. An account can pass a peril loss test and still breach an occupancy-share or regional limit.
Treating a screening factor as exact. Linear incremental loss assumptions can support a preliminary capacity calculation, but they may not capture all portfolio interactions. State the assumption and rerun the portfolio when needed.

The purpose of steering is not to eliminate all exposure to a peril, region, or occupancy. It is to make portfolio changes deliberately, using measures that correspond to the limits and risk appetite being managed. A sound decision records the exposure basis, model output, constraint tested, assumptions, and remaining headroom.

Check Your Understanding

For each hypothetical portfolio, apply the stated assumptions and test all relevant constraints.

A region has $80 million of insured value and a modeled event loss of $12 million. New business adds $10 million of insured value and, under a linear screening assumption, $0.30 of the event loss per dollar insured. The regional event-loss limit is $16 million. What is the maximum addition allowed by this limit, and does the proposed $10 million pass? (Answer: Maximum $13.33 million; the $10 million addition passes, producing $15 million of modeled event loss.)
A portfolio has $80 million total insured value, including $20 million wood-frame exposure. New business is entirely wood-frame, and the wood-frame share must remain at or below 30%. What is the maximum additional wood-frame exposure? (Answer: $5.71 million.)
Three disjoint segments have exposures of $20 million, $40 million, and $40 million, with modeled AALs of $0.12 million, $0.08 million, and $0.10 million, respectively. Find total AAL and the portfolio AAL rate. (Answer: Total AAL is $0.30 million; the portfolio AAL rate is 0.30%.)
Two renewal alternatives are rerun on a complete portfolio. The limits are portfolio 1-in-100 loss at most $30 million, regional 1-in-100 loss at most $18 million, and occupancy share at most 25%. Alternative P reports $29 million, $18 million, and 26%; alternative Q reports $30 million, $17 million, and 24%. Which alternative satisfies every limit? (Answer: Alternative Q only.)
Two segments have modeled 1-in-100 losses of $14 million and $12 million, while a full-portfolio model rerun reports a portfolio 1-in-100 loss of $22 million. Should the segment figures be added to replace the portfolio result? (Answer: No; use the $22 million portfolio result for the portfolio constraint, because segment quantiles generally are not additive.)

What You'll Learn