Allocating Catastrophe Premium Across Risks

Why Allocate a Catastrophe Premium?

A catastrophe model may support a premium indication for an entire portfolio, while pricing and monitoring require amounts for individual locations, policies, or rating groups. An allocation method assigns the portfolio total among those risks. The allocation does not change the total indicated premium; it determines how that total is distributed.

Several reasonable allocation bases can produce different risk-level premiums. Exposure allocation emphasizes the amount insured or another chosen measure of exposure. Average annual loss allocation emphasizes modeled expected losses. Tail-risk allocation emphasizes each risk's contribution in severe portfolio outcomes. The appropriate basis depends on the purpose of the allocation, the available model output, and applicable pricing constraints.

Definition: Suppose a portfolio has $n$ risks and an indicated catastrophe premium $P_{\mathrm{cat}}$. An allocation assigns amounts $P_1,\ldots,P_n$ such that $P_i$ is the catastrophe premium assigned to risk $i$, and a fully reconciled allocation satisfies $\sum_{i=1}^{n}P_i=P_{\mathrm{cat}}$. An allocation share is $w_i=P_i/P_{\mathrm{cat}}$; when the premium is positive, the shares sum to 1.

Allocation is distinct from indication. The portfolio premium may already reflect modeled average annual loss (AAL), a catastrophe risk load, expenses, and profit provisions. The allocation question is how to distribute that indicated amount—not how to re-estimate it. In practice, different components can be allocated using different bases, provided the resulting amounts reconcile to the portfolio total.

Exposure and Average Annual Loss Allocations

Let $E_i$ denote the chosen exposure measure for risk $i$, such as insured value. An exposure allocation assigns premium in proportion to exposure. The method is transparent and gives a common catastrophe premium rate per unit of exposure, but it does not reflect differences in hazard or vulnerability if those are not represented in the exposure measure.

w_i^{(E)}=\frac{E_i}{\sum_{j=1}^{n}E_j}, \qquad P_i=P_{\mathrm{cat}}w_i^{(E)}.

For an AAL allocation, let $X_i$ be the annual catastrophe loss for risk $i$, and let $S=\sum_i X_i$ be portfolio loss. Set $A_i=E[X_i]$. Since expectation is additive, $E[S]=\sum_i A_i$, whether or not the risks are dependent. If the total AAL is positive, an AAL-proportional allocation is

w_i^{(A)}=\frac{A_i}{\sum_{j=1}^{n}A_j}, \qquad P_i=P_{\mathrm{cat}}w_i^{(A)}.

This allocates the entire indicated premium in proportion to expected loss. It is simple, but it may allocate a risk load, expenses, or profit in the same proportions as AAL even when a different basis better reflects those components. A zero-AAL risk receives no premium under this method, even if it has meaningful exposure or tail potential.

Worked Example: Allocating by Insured Exposure

A hypothetical portfolio has an indicated catastrophe premium of $240,000. Three locations have insured exposures of $12 million, $8 million, and $4 million. Allocate the premium in proportion to insured exposure, then compare the premium per $1 million of exposure.

Method. Use the exposure shares. The total exposure is $12+8+4=24$ million dollars.

Location 1 has share $12/24=0.5$, so its allocated premium is $240{,}000(0.5)=\$120{,}000$.

Location 2 has share $8/24=1/3$, so its allocated premium is $240{,}000/3=\$80{,}000$.

Location 3 has share $4/24=1/6$, so its allocated premium is $240{,}000/6=\$40{,}000$.

The premium per $1 million of exposure is $120,000/$12 million = $10,000 for Location 1, $80,000/$8 million = $10,000 for Location 2, and $40,000/$4 million = $10,000 for Location 3. The allocation reconciles: $120{,}000+80{,}000+40{,}000=\$240{,}000$. The common mistake is to treat the allocated amounts as independently indicated premiums; here they are shares of one portfolio total.

Worked Example: Allocating by Average Annual Loss

A hypothetical insurer's indicated catastrophe premium is $154,000. Three rating groups have modeled AALs of $50,000, $30,000, and $20,000. Allocate the premium in proportion to AAL and verify the result.

Method. The total AAL is $50{,}000+30{,}000+20{,}000=\$100{,}000$. Therefore the groups' shares are $50{,}000/100{,}000=0.50$, $30{,}000/100{,}000=0.30$, and $20{,}000/100{,}000=0.20$.

The allocated premiums are $154{,}000(0.50)=\$77{,}000$, $154{,}000(0.30)=\$46{,}200$, and $154{,}000(0.20)=\$30{,}800$.

The check is $77{,}000+46{,}200+30{,}800=\$154{,}000$, and the shares sum to $0.50+0.30+0.20=1$. The method allocates the indicated premium, not merely the $100,000 of AAL. A common error is to allocate only the AAL dollars and overlook that the premium being distributed is $154,000.

Allocating a Tail-Risk Load

A portfolio catastrophe premium may include a load for uncertainty or severe loss in addition to AAL. Allocating that load using AAL alone may not reflect which risks drive portfolio tail outcomes. One practical approach uses component losses in the portfolio's tail scenarios. For a continuous aggregate loss $S$, a tail contribution at confidence level $q$ can be defined as the conditional expected component loss $E[X_i\mid S>\operatorname{VaR}_q(S)]$, where $\operatorname{VaR}_q(S)$ is the $q$-quantile of aggregate loss. The component contributions sum to the corresponding conditional expected aggregate loss because $S=\sum_iX_i$ and conditional expectation is additive.

For a discrete simulated distribution, the event $S>\operatorname{VaR}_q(S)$ may not select exactly the upper $1-q$ fraction of outcomes. A consistent tail convention is therefore essential. One standard definition is TVaR (also called expected shortfall), the average of the quantile function over probabilities from $q$ to 1; in discrete cases this definition accounts for the appropriate fraction of probability at the quantile. Component contributions must be calculated using that same convention if they are to reconcile to portfolio TVaR.

Worked Example: Allocating a Premium by Tail Contribution

A hypothetical two-risk portfolio has four equally likely annual outcomes. Losses, in thousands of dollars, are: $(X_1,X_2)=(0,0),(100,0),(0,150),(200,100)$. Allocate an indicated catastrophe premium of $180,000 in proportion to the risks' contributions to 75% TVaR.

Method. The aggregate losses in the four outcomes are $0,100,150,300$ thousand dollars. Each outcome has probability 0.25. The 75th percentile is $150,000, and the upper-quarter tail consists of the $300,000 outcome. Thus 75% TVaR is $300,000. In that tail outcome, Risk 1 contributes $200,000 and Risk 2 contributes $100,000. Their tail contributions sum to the portfolio tail loss.

The contribution shares are $200{,}000/300{,}000=2/3$ for Risk 1 and $100{,}000/300{,}000=1/3$ for Risk 2. The allocated premiums are $180{,}000(2/3)=\$120{,}000$ and $180{,}000(1/3)=\$60{,}000$.

The allocation check is $120{,}000+60{,}000=\$180{,}000$. For comparison, the AALs are $75{,}000$ and $62{,}500$, respectively, so an AAL allocation would assign about $98,181.82 and $81,818.18. The difference is not an arithmetic error: the tail allocation gives more weight to the risk that contributes more to the severe portfolio outcome. The common mistake is to use a component's stand-alone worst loss as though it were its contribution to the portfolio tail.

Why Marginal Contributions Need Care

A tempting tail allocation is to calculate, for each risk, the change in a portfolio risk measure when that risk is removed. If $\rho(S)$ is a risk measure, this difference is $\rho(S)-\rho(S-X_i)$. Such a marginal amount can be informative, but the amounts generally do not sum to $\rho(S)$. Removing one risk at a time changes the portfolio context, and diversification or dependence can make those changes interact.

A related allocation uses Euler contributions when the risk measure is differentiable and homogeneous of degree one in the scaled risks. Let $x_i$ be the scale of risk $i$, and write the portfolio risk measure as $\rho(x_1,\ldots,x_n)$. Homogeneity means $\rho(cx_1,\ldots,cx_n)=c\rho(x_1,\ldots,x_n)$ for $c>0$. Under differentiability, Euler's theorem gives the allocation below.

\sum_{i=1}^{n}x_i\frac{\partial \rho}{\partial x_i}=\rho.

To see why, define $g(c)=\rho(cx_1,\ldots,cx_n)$. Homogeneity gives $g(c)=c\rho(x_1,\ldots,x_n)$, so $g'(1)=\rho(x_1,\ldots,x_n)$. The chain rule also gives $g'(1)=\sum_i x_i(\partial\rho/\partial x_i)$. Equating the two expressions proves the result. Thus these Euler contributions reconcile to the portfolio risk measure, subject to the stated assumptions. If the measure is not differentiable, an appropriate subgradient or another explicitly defined allocation convention may be needed; Euler allocation should not be applied mechanically.

Comparing Relativities and Checking the Allocation

An allocation can be converted to a rate or relativity only after the denominator is specified. For example, if $E_i$ is insured exposure, $P_i/E_i$ is the allocated catastrophe premium per unit of exposure. A relativity to a reference risk is the ratio of their rates, not necessarily the ratio of their allocated premium dollars. Risks with different exposure amounts can have very different allocated premiums but the same rate.

Before using allocations in a rating plan, check the underlying basis and the reconciliation:

Confirm that every risk is included once and that the chosen exposure or loss measure uses consistent units.
Check that the shares sum to 1 and that allocated premiums sum to the portfolio indication.
For tail allocations, state the tail measure, confidence level, and convention for discrete outcomes.
Compare the resulting rates or relativities with the purpose of the allocation; a mathematically reconciled allocation is not automatically an appropriate rating decision.
If some contributions are negative or the chosen weights sum to zero, a direct proportional allocation may be unsuitable. Investigate the cause and select a defensible alternative rather than silently forcing the weights to work.

A unit error can overwhelm an otherwise sound method. Suppose a hypothetical portfolio provision is $120,000 and an additional charge is $30,000 per exposure unit across 1,000 units. The total is

120{,}000+30{,}000(1{,}000)=120{,}000+30{,}000{,}000 =\$30{,}120{,}000.

It is not $150,000: that incorrect result adds the $30,000 charge once, rather than multiplying the per-unit amount by 1,000. If the resulting $30,120,000 were allocated equally across the 1,000 units, each would receive $30{,}120{,}000/1{,}000=\$30{,}120$, and the allocations would again sum to the portfolio amount. Always distinguish a portfolio total from an amount stated per location or exposure unit.

Key takeaway: Choose an allocation basis that matches the component being distributed, calculate shares on a consistent basis, and verify that the allocated dollars reconcile to the indicated portfolio premium. Exposure, AAL, and tail-risk allocations answer different questions and need not produce the same relativities.

Check Your Understanding

For each problem, identify the allocation basis and check that the allocated amounts reconcile.

A hypothetical $90,000 catastrophe premium is allocated by exposures of $6 million, $3 million, and $1 million. What premium is assigned to each risk? (Answer: $54,000, $27,000, and $9,000.)
A $100,000 indicated premium is allocated among three groups with AALs of $18,000, $12,000, and $10,000. What are the allocations? (Answer: $45,000, $30,000, and $25,000.)
Four equally likely outcomes have risk losses $(0,0),(80,20),(20,100),(160,40)$, in thousands of dollars. Using the upper-quarter outcome as the 75% TVaR tail, allocate a $250,000 premium in proportion to tail contributions. (Answer: Risk 1 receives $200,000 and Risk 2 receives $50,000.)
A $240,000 premium has a $120,000 AAL-based component and a $120,000 tail-load component. The AALs for A, B, and C are $30,000, $20,000, and $10,000; their tail contributions are $40,000, $40,000, and $20,000. Allocate each component on its stated basis and combine them. (Answer: A receives $108,000, B receives $88,000, and C receives $44,000.)
A portfolio provision is $85,000 plus $12,500 per insured location for 800 locations. Find the total and, if allocated equally, the amount per location. (Answer: Total $10,085,000; $12,606.25 per location.)

What You'll Learn