Why Capital Relief Requires More Than a Payment Estimate
An alternative risk transfer (ART) transaction may reduce an insurer’s exposure to adverse losses, provide funds after an event, or do both. Those effects are not interchangeable. A payment that must be repaid can improve liquidity without reducing the insurer’s ultimate loss, while a transfer that reduces losses may provide less capital relief than expected if recovery is uncertain or the capital framework limits recognition.
A sound comparison therefore starts with the loss distribution the insurer retains after the transaction. It then applies the same capital measure to the gross and retained distributions, and separately considers the cost of financing, collateral collectability, and any recognition limits. The distinction is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
For a confidence level \(\alpha\), one common economic capital measure is unexpected-loss capital:
Here \(\operatorname{TVaR}_\alpha(X)\) is the average loss in the worst \(1-\alpha\) fraction of outcomes. Under this definition, economic capital relief is
A positive value indicates lower modeled unexpected-loss capital after the transaction. A negative value means the retained distribution has greater capital under the chosen measure. This calculation assumes that gross and retained losses are modeled on the same basis, over the same horizon, and at the same confidence level. If another capital measure or a different risk horizon is required, it must be used consistently for both distributions.
Why the Tail Calculation Works
For a loss \(W\), define \(\operatorname{VaR}_u(W)\) as the \(u\)-quantile of its distribution. TVaR can be written as the average of quantiles over the upper tail:
The formula averages a tail of total probability \(1-\alpha\), rather than simply averaging outcomes strictly above the quantile. This distinction matters for a discrete loss distribution: the probability mass at the quantile can be larger than the remaining tail probability, so only part of that mass belongs in the worst \(1-\alpha\) outcomes.
To see why the integral gives the correct tail average, imagine dividing the probability scale from \(\alpha\) to 1 into small intervals. Each interval has probability width \(du\), and its quantile gives the loss associated with that part of the distribution. Summing loss times probability across those intervals gives the total loss in the upper tail; dividing by the tail probability \(1-\alpha\) gives its average. For a discrete distribution, the same reasoning means including all worse outcomes first and taking only the necessary fraction of the boundary outcome.
Worked Example: Discrete Retained Loss and Capital Relief
Worked Example: Discrete Retained Loss and Capital Relief
A hypothetical insurer evaluates an ART arrangement using a 95% TVaR capital measure. Its model gives gross unexpected-loss capital of $88.20 thousand. After allowing for the transaction’s recovery and collection assumptions, the retained loss \(Z\), in thousands of dollars, has this distribution:
| Retained loss \(Z\) | Probability |
|---|---|
| 0 | 0.8993 |
| 20 | 0.0707 |
| 80 | 0.0294 |
| 90 | 0.0006 |
Find retained-loss capital and the modeled capital relief. State how the worst 5% is selected.
Method. Calculate \(E[Z]\), then average exactly the worst 5% of retained-loss outcomes. The gross capital is supplied as a model result; it is not recalculated from the retained-loss distribution.
Step 1: Check the probabilities and calculate the mean. The probabilities sum to \(0.8993+0.0707+0.0294+0.0006=1.0000\). Thus,
The mean retained loss is $3.820 thousand.
Step 2: Select the worst 5%. Start with the greatest losses: probability \(0.0006\) at 90, followed by \(0.0294\) at 80. These account for \(0.0300\) of the required \(0.0500\) tail probability. The remaining amount is \(0.0500-0.0300=0.0200\), taken from the probability mass at 20.
Step 3: Calculate TVaR and capital.
Therefore, \(K_{0.95}(Z)=56.12-3.82=52.30\) thousand dollars. Modeled capital relief is \(88.20-52.30=35.90\) thousand dollars.
Check. The selected tail probability is \(0.0006+0.0294+0.0200=0.0500\). Its average, 56.12, lies between the smallest and largest losses selected, 20 and 90. The common mistake is to take only part of the probability at 80 and then add too much probability at 20. That would replace worse outcomes with better ones, so it would not be the worst 5% of outcomes.
Collateral and Collection Constraints Change the Loss Distribution
A contractual recovery is not necessarily a collectible recovery. If collateral is subject to a haircut, or a payment is reduced by a recovery constraint, the retained-loss model should reflect that before capital is calculated. The haircut changes the loss distribution itself; it is not simply a fee to subtract after calculating capital relief.
Worked Example: A Collateral Haircut and a Recognition Limit
A hypothetical insurer has gross losses of $0 thousand with probability 0.90, $60 thousand with probability 0.08, and $140 thousand with probability 0.02. A cover has a $40 thousand attachment and an $80 thousand maximum payment. The amount collectible is only 90% of the contractual payment because of a collateral haircut. Calculate gross and retained unexpected-loss capital at 95% TVaR, then calculate recognized relief if the applicable capital rule recognizes only 75% of modeled relief.
Method. For gross loss \(X\), the contractual payment is \(\min(\max(X-40,0),80)\). Apply the 90% collectability factor to that payment, then subtract the collectible amount from gross loss.
Step 1: Calculate collectible payments and retained losses. At \(X=0\), the contractual and collectible payments are both 0, leaving \(Z=0\). At \(X=60\), the contractual payment is \(60-40=20\), the collectible payment is \(0.90(20)=18\), and \(Z=60-18=42\). At \(X=140\), the contractual payment is capped at 80, the collectible payment is \(0.90(80)=72\), and \(Z=140-72=68\).
Step 2: Calculate gross capital.
The worst 5% includes all 2% of the $140 thousand outcome and 3% of the 8% probability at $60 thousand.
Step 3: Calculate retained capital.
Step 4: Separate modeled and recognized relief. Modeled relief is \(84.4-47.68=36.72\) thousand dollars. Recognized relief under the stated 75% limit is \(0.75(36.72)=27.54\) thousand dollars. The remaining difference is not an additional loss in this model; it is relief that the stated rule does not recognize.
Check. The retained capital is below gross capital, consistent with the collectible payments reducing retained losses. The 75% recognition factor is applied to modeled relief, not to gross capital or to each individual loss outcome.
Financing Can Improve Liquidity Without Reducing Loss
A financing feature must be modeled according to its economic effect. A repayable drawdown may provide cash when losses occur, but the insurer still owes the amount drawn. If the drawdown does not reduce or transfer the insurer’s loss obligation, it is not a recovery \(R\) in \(Z=X-R\). A fixed commitment fee may reduce earnings or available resources, but by itself it does not change an unexpected-loss measure defined as TVaR minus expected loss.
Worked Example: Contingent Financing Versus Loss Transfer
A hypothetical insurer has gross loss \(X\), in thousands of dollars, with probability 0.90 at 0, probability 0.08 at 80, and probability 0.02 at 200. A contingent financing facility can advance $40 thousand after a severe event, but the advance is repayable and does not satisfy any claim obligation. The facility also charges a fixed $1.20 thousand fee. Compare this arrangement with a hypothetical indemnity cover that pays the loss layer from 100 to 200, with no collection shortfall or premium included in this capital calculation.
Step 1: Calculate gross capital.
Step 2: Assess the financing facility. Because the advance is repayable and does not discharge claims, retained loss remains \(Z=X\). Thus, unexpected-loss capital remains $117.60 thousand under the stated measure. The $1.20 thousand fixed fee affects the insurer’s financial position, but it does not change this particular capital calculation: subtracting the same fixed amount from every outcome reduces both TVaR and the mean by $1.20 thousand, leaving their difference unchanged. The repayment obligation must still be included in a broader balance-sheet or liquidity assessment.
Step 3: Assess the indemnity cover. The layer payment is \(\min(\max(X-100,0),100)\). It is zero at losses 0 and 80, and 100 at loss 200. Retained losses are therefore 0, 80, and 100 at the three respective outcomes.
Modeled capital relief from the indemnity cover is \(117.6-79.6=38.0\) thousand dollars. The financing facility may help meet immediate cash needs, but it produces no reduction in modeled unexpected-loss capital in this example.
Check. The indemnity reduces losses in the highest outcome and therefore lowers the upper-tail average. The financing advance is not incorrectly counted as a claim recovery simply because cash becomes available.
What to Include in a Capital-Relief Comparison
Capital relief is a comparison of modeled risks, not a statement that every economic benefit of a transaction is captured in one number. Keep the principal components distinct:
| Component | How to treat it | What it does not establish by itself |
|---|---|---|
| Collectible risk transfer | Use the collectible payment when constructing retained loss | That every contractual payment will be collected |
| Financing proceeds | Model cash availability and repayment obligations separately | That the insurer’s ultimate loss has been reduced |
| Collateral haircut or shortfall | Reduce modeled recovery before calculating retained capital | That nominal collateral equals usable recovery |
| Transaction fees | Include in earnings, available resources, or a total-cost analysis as appropriate | That a fixed fee changes unexpected-loss capital as defined here |
| Recognition limit | Apply the stated limit to modeled relief when calculating recognized relief | That unrecognized modeled relief is an additional claim payment |
Basis risk can also affect the relationship between a trigger and the insurer’s actual loss. If the transaction pays according to an index or other measure that does not exactly track the insurer’s loss, the payment distribution must be modeled jointly with the loss. Treating an expected payment as though it were certain in every relevant loss scenario can overstate relief. Likewise, a recovery constraint should be reflected in the scenario outcomes rather than appended as an unexplained percentage after the calculation.
Common Mistakes and Exam Traps
- Using the wrong tail for a discrete distribution. Include the worst \(1-\alpha\) probability, not an arbitrary collection of outcomes near the quantile. A boundary outcome may contribute only part of its probability mass.
- Comparing inconsistent quantities. Gross and retained capital must use the same horizon, confidence level, and capital definition. A TVaR result cannot be subtracted from a VaR result and interpreted as relief.
- Counting a loan as a recovery. Cash advances that must be repaid do not necessarily reduce claims or retained losses. Model their liquidity and repayment effects separately.
- Ignoring collection limits. A nominal payment subject to collateral haircuts or counterparty nonperformance is not automatically the amount collected. Model the effective recovery.
- Confusing modeled relief with recognized relief. A capital framework may recognize only part of the modeled reduction. State both amounts and the rule connecting them.
Check Your Understanding
For each problem, calculate the requested capital quantity using the stated unexpected-loss measure unless the question asks for another result. Amounts are in thousands of dollars.
- A gross loss is 0 with probability 0.90, 40 with probability 0.07, and 100 with probability 0.03. A certain cover pays \(\min(X,50)\). Find gross and retained 95% unexpected-loss capital and modeled relief. (Answer: Gross capital 70.20; retained capital 28.50; relief 41.70.)
- A gross loss is 0 with probability 0.95, 100 with probability 0.04, and 200 with probability 0.01. The transaction pays 80% of \(\min(X,100)\), with no other collection risk. Find gross and retained 95% unexpected-loss capital and modeled relief. (Answer: Gross capital 114; retained capital 38; relief 76.)
- A repayable facility advances 40 after a severe loss but does not discharge claims. Gross capital under the stated measure is 117.60. A fixed facility fee is 1.20. What is the retained unexpected-loss capital under this measure, and what is the fee’s effect on this capital measure? (Answer: Retained capital remains 117.60; the fixed fee does not change TVaR minus expected loss, although it reduces financial resources by 1.20.)
- A retained loss is 0 with probability 0.94, 50 with probability 0.04, and 150 with probability 0.02. Calculate 95% TVaR and unexpected-loss capital, showing the tail allocation. (Answer: The worst 5% contains 0.02 at 150 and 0.03 at 50; TVaR is 90 and capital is 85.)
- A transaction reduces modeled unexpected-loss capital from 52 to 26. A capital rule recognizes 60% of modeled relief. Find modeled and recognized relief, and state how much modeled relief remains unrecognized. (Answer: Modeled relief is 26; recognized relief is 15.6; unrecognized relief is 10.4.)