Translating Capital Targets into Risk Loads

From Capital Requirement to Premium

A capital target is not automatically a premium load. Capital is an amount held to support risk; a premium load is an amount collected to pay for the cost of holding that capital. The conversion requires a pricing rule, such as a cost-of-capital rate. It also requires the loss, capital, expenses, and premium to be measured on compatible bases: for example, all for the same annual period and the same portfolio.

The distinction matters because two portfolios with the same expected loss may have different capital needs, and two insurers facing the same capital need may apply different required returns. The risk measure determines the capital amount; the pricing rule determines how much of that amount enters the premium.

Definition: Let $X$ be the aggregate loss for a stated portfolio and period, with finite expected value $\mu=E[X]$. If $\rho(X)$ is the selected capital risk measure, define the capital requirement above expected loss as $K=\rho(X)-\mu$. For a cost-of-capital rate $c$ for that period, the capital cost included in the premium is $cK$. Unless a specified pricing rule floors capital at zero, use the calculated $K$ as given; do not silently replace a negative value with zero.

This convention makes the capital amount incremental to the expected loss: expected losses are already funded through the pure premium. If a capital target is supplied directly as an amount above expected loss, use that amount as $K$. If the target is instead a total risk measure $\rho(X)$, subtract expected loss before applying the cost-of-capital rate.

Three Ways to Define a Capital Add-On

A variance-based rule, a VaR rule, and a TVaR rule can produce different capital amounts for the same loss distribution. Their units and interpretations differ, so the capital definition must be stated before calculating the premium load.

Variance-based capital: A calibrated rule can define $K_{\mathrm{var}}=\gamma\operatorname{Var}(X)$, where $\gamma$ has units inverse to those of $X$. Variance alone is measured in squared currency units, so it cannot be added directly to a currency amount. The coefficient $\gamma$ supplies the necessary scaling; there is no universal value for it.

L_{\mathrm{var}}=cK_{\mathrm{var}} =c\gamma\operatorname{Var}(X)

Here $L_{\mathrm{var}}$ is the premium amount allocated to the cost of capital. A variance rule is a pricing choice, not a translation-invariant risk measure: adding a fixed amount to every loss changes the mean but not the variance. Its coefficient and intended use must therefore be clear.

Definition: For confidence level $p$, define $\operatorname{VaR}_p(X)=\inf\{x:F_X(x)\ge p\}$. The VaR-based capital above expected loss is $K_{\mathrm{VaR}}=\operatorname{VaR}_p(X)-E[X]$. Define $\operatorname{TVaR}_p(X)$ as the average of the upper $1-p$ fraction of the loss distribution, equivalently $\operatorname{TVaR}_p(X)=\frac{1}{1-p}\int_p^1\operatorname{VaR}_u(X)\,du$ when the integral exists. Then $K_{\mathrm{TVaR}}=\operatorname{TVaR}_p(X)-E[X]$.

For continuous distributions, TVaR can often be calculated as the conditional expected loss above the VaR threshold. That shortcut needs care for discrete losses: probability mass at the threshold may have to be included only in part. The quantile-integral definition handles both continuous and discrete distributions.

Capital basis	Capital above expected loss	Capital cost in premium
Calibrated variance	$\gamma\operatorname{Var}(X)$	$c\gamma\operatorname{Var}(X)$
VaR at level $p$	$\operatorname{VaR}_p(X)-E[X]$	$c[\operatorname{VaR}_p(X)-E[X]]$
TVaR at level $p$	$\operatorname{TVaR}_p(X)-E[X]$	$c[\operatorname{TVaR}_p(X)-E[X]]$

How the Premium Equation Follows

Suppose all amounts relate to the same portfolio and annual period. Let $P$ be the premium, let $q$ be a proportional expense rate applied to premium, let $\mu=E[X]$, and let $K$ be the capital requirement above expected loss. Assume the pricing rule collects expected loss, proportional expenses, and a capital cost equal to $cK$. This is a simplified pricing balance: it does not include other fixed expenses, taxes, investment income, or discounting.

Identify the cost items.
Expected loss is $\mu$; the proportional expense amount is $qP$; and the capital cost is $cK$.

Write the balance.
The premium funds all three items, so $P=\mu+qP+cK$.

Solve for premium.
Rearranging gives $(1-q)P=\mu+cK$, hence $P=(\mu+cK)/(1-q)$, provided $q<1$.

The equation shows why the capital cost should not be confused with the total premium increase when expenses are a percentage of premium. Raising the numerator by $cK$ also raises proportional expenses, so the gross premium rises by $cK/(1-q)$, not merely $cK$. If expenses are not proportional, use the actual expense formula rather than this gross-up.

Worked Applications

Worked Example: A Calibrated Variance-Based Capital Target

A hypothetical annual portfolio has expected aggregate losses of $500$ thousand dollars and loss variance of $2{,}500$ $(\text{thousand dollars})^2$. The pricing rule defines capital as $K_{\mathrm{var}}=\gamma\operatorname{Var}(X)$, with $\gamma=0.04$ per thousand dollars. The cost-of-capital rate is $c=10\%$, and proportional expenses are $q=5\%$ of premium. Calculate the premium. All amounts below are in thousands of dollars.

Method. Calculate variance-based capital, apply the cost-of-capital rate, and then gross up for proportional expenses.

First calculate capital:

K_{\mathrm{var}} =0.04\frac{1}{\text{thousand dollars}} \times 2{,}500(\text{thousand dollars})^2 =100\text{ thousand dollars}.

The units reduce to thousands of dollars, as required. The capital cost is

cK_{\mathrm{var}}=0.10(100)=10.

Now solve the premium equation:

P=\frac{500+10}{1-0.05} =\frac{510}{0.95} =536.842105\ldots

To check the balance, proportional expenses are $0.05(536.842105\ldots)=26.842105\ldots$. Expected losses, expenses, and capital cost sum to $500+26.842105\ldots+10=536.842105\ldots$, the premium. Rounded to three decimals in thousands, the premium is $536.842$ thousand dollars.

Common mistake avoided: Adding the capital amount $100$ directly to the expected loss would give $600$ before expenses. That treats the capital requirement itself as a premium charge. Under the stated rule, the charge is $10$, because the cost-of-capital rate is $10\%$.

Worked Example: Comparing VaR and TVaR Capital on a Discrete Loss

A hypothetical annual aggregate loss $X$, measured in thousands of dollars, is $0$ with probability $0.50$, $1{,}000$ with probability $0.40$, and $3{,}000$ with probability $0.10$. Compare the premium capital costs at $p=0.90$, using a cost-of-capital rate of $12\%$. Ignore expenses.

Method. Calculate expected loss, then calculate each capital measure above expected loss before applying the same cost-of-capital rate.

Expected loss is

$$ E[X]=0(0.50)+1{,}000(0.40)+3{,}000(0.10)=700. $$

The cumulative probabilities are $F_X(0)=0.50$, $F_X(1{,}000)=0.90$, and $F_X(3{,}000)=1.00$. By the stated lower-quantile definition, $\operatorname{VaR}_{0.90}(X)=1{,}000$. Thus

K_{\mathrm{VaR}}=1{,}000-700=300, \qquad L_{\mathrm{VaR}}=0.12(300)=36.

For TVaR, the upper $10\%$ of the probability distribution consists exactly of the $3{,}000$ outcome, which has probability $0.10$. Therefore

\operatorname{TVaR}_{0.90}(X)=3{,}000, \qquad K_{\mathrm{TVaR}}=3{,}000-700=2{,}300, \qquad L_{\mathrm{TVaR}}=0.12(2{,}300)=276.

With no expenses, the corresponding premiums are $700+36=736$ thousand dollars under VaR and $700+276=976$ thousand dollars under TVaR. The values differ because VaR records the threshold loss, whereas TVaR averages the upper tail.

Common mistake avoided: Using $E[X\mid X>1{,}000]$ for TVaR would give $3{,}000$ in this example, but that conditional-expectation shortcut is not generally valid for discrete losses with mass at the VaR threshold. Here the upper $10\%$ happens to be exactly the $3{,}000$ outcome; the quantile-integral definition is the reliable general method.

Worked Example: Keeping Portfolio and Per-Policy Bases Consistent

A hypothetical annual book of $4{,}000$ policies has expected aggregate losses of $1{,}200{,}000$ dollars. A stated risk measure implies capital of $240{,}000$ dollars above expected loss for that same annual book. The cost-of-capital rate is $10\%$, and proportional expenses are $8\%$ of premium. Calculate the aggregate premium and the premium per policy, assuming equal allocation across policies.

Method. Treat the supplied capital as an aggregate capital amount above expected aggregate loss, calculate its cost, and apply the expense gross-up on the aggregate premium.

The capital cost for the book is

cK=0.10(240{,}000)=24{,}000\text{ dollars}.

The aggregate premium is

P=\frac{1{,}200{,}000+24{,}000}{1-0.08} =\frac{1{,}224{,}000}{0.92} =1{,}330{,}434.7826\ldots\text{ dollars}.

Dividing by $4{,}000$ policies gives

\frac{1{,}330{,}434.7826\ldots}{4{,}000} =332.6086957\ldots\text{ dollars per policy}.

The aggregate expense amount is $0.08(1{,}330{,}434.7826\ldots)=106{,}434.7826\ldots$ dollars. Adding this to expected losses and capital cost gives $1{,}200{,}000+106{,}434.7826\ldots+24{,}000=1{,}330{,}434.7826\ldots$, which checks. Rounded to the nearest cent, the premium is $1{,}330{,}434.78$ dollars in total, or $332.61$ dollars per policy.

Common mistake avoided: Dividing aggregate expected loss by the number of policies but leaving the capital amount at the aggregate level mixes bases. Either calculate the whole book first and divide the resulting premium by policy count, as above, or allocate both expected loss and capital consistently per policy.

Basis Checks and Exam Traps

Before accepting a risk load, verify what each input represents. A correct calculation on mismatched bases can still produce a misleading indication.

Match the period. An annual cost-of-capital rate must be applied to capital for the corresponding annual period. Do not multiply a monthly capital amount by an annual rate without a consistent annualization rule.
Match the portfolio. If the target capital is for a whole book, use whole-book expected losses and premium, or allocate all components consistently to a policy or exposure unit.
Match the loss basis. Use losses and capital measured on the same gross or net basis. For example, do not combine net expected losses with a capital target calculated on gross losses unless the pricing rule explicitly calls for it.
Subtract expected loss only once. If the supplied target is already capital above expected loss, it is $K$. Subtracting expected loss again understates the capital cost. If the target is a total measure $\rho(X)$, then use $K=\rho(X)-E[X]$.
Do not confuse a capital amount with its price. The capital requirement is $K$; under a cost-of-capital rule, the premium charge is $cK$, with any required expense gross-up applied afterward.
Check units. VaR and TVaR are currency amounts. Variance is squared currency, so a variance-based rule needs a coefficient with inverse-currency units before it can define a currency capital amount.

These checks are especially useful when a problem gives a capital target and a premium base in different forms. State whether each number is per policy, per exposure unit, or aggregate, and whether it is before or after expenses. Then write the pricing balance in those same units before substituting values. This method of translating a capital target into a premium risk load is aligned with the SOA Exam ASTAM (Advanced Short-Term Actuarial Mathematics) syllabus.

Check Your Understanding

For each problem, identify the capital cost separately from the total premium. Unless stated otherwise, amounts refer to the same annual portfolio and ignore expenses.

A hypothetical book has expected loss of $240$ thousand dollars and variance of $900$ $(\text{thousand dollars})^2$. Capital is defined as $K=0.10\operatorname{Var}(X)$, with the coefficient in inverse thousands of dollars. If the cost-of-capital rate is $8\%$, find the capital and its premium cost. (Answer: $K=90$ thousand dollars; capital cost $=7.2$ thousand dollars.)
A hypothetical loss has outcomes $0$, $500$, and $1{,}500$ thousand dollars with probabilities $0.60$, $0.30$, and $0.10$. At $p=0.90$, find expected loss, VaR capital, and the capital cost at $10\%$. (Answer: $E[X]=300$ thousand dollars; $\operatorname{VaR}_{0.90}=500$ thousand dollars; capital $=200$ thousand dollars; capital cost $=20$ thousand dollars.)
A loss has expected value $800$ thousand dollars and $\operatorname{TVaR}_{0.95}=1{,}300$ thousand dollars. With cost of capital $12\%$ and no expenses, find capital, capital cost, and premium. (Answer: capital $=500$ thousand dollars; capital cost $=60$ thousand dollars; premium $=860$ thousand dollars.)
A hypothetical annual portfolio has expected losses of $900{,}000$ dollars, capital of $150{,}000$ dollars above expected losses, cost of capital $10\%$, and expenses of $10\%$ of premium. Find aggregate premium. (Answer: $1{,}016{,}666.67$ dollars.)
A book of $2{,}500$ policies has expected aggregate losses of $750{,}000$ dollars and aggregate capital of $100{,}000$ dollars above expected loss. The cost-of-capital rate is $15\%$; expenses are $5\%$ of premium. Find the premium per policy under equal allocation. (Answer: aggregate premium $=805{,}263.16$ dollars; premium per policy $=\$322.11$.)

Capital basis	Capital above expected loss	Capital cost in premium
Calibrated variance	\(\gamma\operatorname{Var}(X)\)	\(c\gamma\operatorname{Var}(X)\)
VaR at level \(p\)	\(\operatorname{VaR}_p(X)-E[X]\)	\(c[\operatorname{VaR}_p(X)-E[X]]\)
TVaR at level \(p\)	\(\operatorname{TVaR}_p(X)-E[X]\)	\(c[\operatorname{TVaR}_p(X)-E[X]]\)

What You'll Learn