Why Concentration and Dependence Matter
In a hypothetical large-deductible program, an insurer may pay covered claims and then seek reimbursement from the insured. Cash, a trust, or a letter of credit may provide security for that obligation. The security analysis is not only about how much reimbursement is expected. It also asks how much could be unpaid at the same time and whether the event producing a large obligation also weakens the insured or the security.
Two features can make a security position more vulnerable. Concentration means that a material share of exposure depends on one counterparty, asset, industry, or other common source of risk. Wrong-way risk arises when the amount at risk increases, or the value of available recovery decreases, as the counterparty becomes more likely to fail. For example, a severe operating event could generate unusually large deductible reimbursements while also reducing the insured’s liquidity and the value of collateral tied to its business.
These ideas complement the earlier material on measuring unpaid deductible obligations and counterparty credit risk. They do not replace claim-level measurement: the relevant amount is still the unpaid obligation after accounting for applicable security and recoveries. The added step is to assess the joint conditions under which the obligation, default, and collateral shortfall occur.
Concentration: Expected Loss and Tail Loss Are Different
For a portfolio of counterparties, total security loss is the sum of the losses for each counterparty. Linearity of expectation means that dependence between counterparties does not, by itself, change the expected total if each counterparty’s expected loss remains the same. Dependence can, however, change the distribution of the total: a common stress can make simultaneous large losses much more likely.
A concentrated position can therefore have a similar expected loss to a diversified position but a larger potential loss from one event. Diversification is not a guarantee of lower expected loss; it can reduce the chance that a single counterparty failure creates a large aggregate shortfall, provided exposures do not share a strong common driver.
Worked Example: One Large Counterparty Versus Two Smaller Ones
A hypothetical insurer is owed $100,000 under one large-deductible account. It holds $60,000 of recoverable collateral, and the insured has a 10% probability of default during the assessment period. Compare this with two separate insureds, each owing $50,000 and each supported by $30,000 of recoverable collateral. Each of the two insureds has a 10% default probability, and their defaults are independent. Calculate expected shortfall in both arrangements and the probability that total shortfall is at least $40,000.
Method. For each account, calculate the shortfall conditional on default as owed amount less recoverable collateral, floored at zero. Multiply by default probability for expected loss.
For the single account, the shortfall if default occurs is \(100{,}000-60{,}000=\$40{,}000\). Its expected shortfall is \(0.10(\$40{,}000)=\$4{,}000\). Total shortfall is at least $40,000 exactly when this insured defaults, so that probability is 10%.
For each of the two smaller accounts, the shortfall on default is \(\$50{,}000-\$30{,}000=\$20{,}000\). Each account has expected shortfall \(0.10(\$20{,}000)=\$2{,}000\), so combined expected shortfall is \(\$2{,}000+\$2{,}000=\$4{,}000\).
With independent defaults, combined shortfall reaches $40,000 only if both insureds default. Thus the probability is \(0.10(0.10)=1\%\). The expected shortfall is the same $4,000 in both arrangements, but the probability of a shortfall of at least $40,000 is 10% for the concentrated account and 1% for the independent pair. Check: in the two-account arrangement, one default causes $20,000 of shortfall and two cause $40,000, matching the threshold calculation. The common mistake is to infer that equal expected losses imply equal tail risk; they do not.
Wrong-Way Risk: Keep the Joint Outcome
A useful way to formalize wrong-way risk is to keep default and potential shortfall together in the calculation. Since \(D\) is an indicator,
This identity follows by conditioning on whether default occurs: \(DU=0\) when \(D=0\), and \(DU=U\) when \(D=1\). Under independence between default and \(U\), the expression reduces to \(P(D=1)E[U]\). If default is more likely in states where \(U\) is high, then \(E[U\mid D=1]\) exceeds the unconditional average and the independence calculation understates expected loss.
The same point can be expressed with covariance. Because \(D\) is an indicator, \(E[D]=P(D=1)\), and
This follows directly from the definition \(\operatorname{Cov}(D,U)=E[DU]-E[D]E[U]\). Positive covariance means the joint loss exceeds the product-of-averages estimate. A negative covariance would reduce expected loss relative to that estimate, although a negative relationship should not be assumed without support.
Worked Example: Default Becomes More Likely as Collateral Weakens
A hypothetical account has $100,000 owed in either of two economic states. In the normal state, which has probability 0.80, default probability is 0.02 and recoverable collateral is $90,000. In the stress state, which has probability 0.20, default probability is 0.40 and recoverable collateral is $50,000. Calculate expected security shortfall using the state-specific assumptions. Then compare it with an estimate that incorrectly treats default and collateral value as independent.
State-specific calculation. In the normal state, the shortfall on default is \(\$100{,}000-\$90{,}000=\$10{,}000\). The contribution to expected shortfall is \(0.80(0.02)(\$10{,}000)=\$160\).
In the stress state, the shortfall on default is \(\$100{,}000-\$50{,}000=\$50{,}000\). The contribution is \(0.20(0.40)(\$50{,}000)=\$4{,}000\). Therefore expected shortfall is \(\$160+\$4{,}000=\$4{,}160\).
Independence estimate. The unconditional default probability is \(0.80(0.02)+0.20(0.40)=0.016+0.080=0.096\). The unconditional average potential shortfall is \(0.80(\$10{,}000)+0.20(\$50{,}000)=\$8{,}000+\$10{,}000=\$18{,}000\). Multiplying these averages gives \(0.096(\$18{,}000)=\$1{,}728\), less than the joint-state result of $4,160.
The discrepancy occurs because the state with weaker collateral also has the higher default probability. Check: the state probabilities sum to 1, and the contributions use the probability of the state, conditional default probability, and loss conditional on default. The common mistake is to multiply averages when the model specifies material dependence between default and shortfall.
Scenario Analysis for Joint Claim, Credit, and Collateral Stress
A practical scenario model can represent several drivers at once. For each scenario, specify its probability, the unpaid deductible obligation, the conditional default probability, and recoverable collateral. Calculate the conditional shortfall \(U=\max(O-C,0)\), then multiply it by both the scenario probability and conditional default probability. The result is a scenario contribution to expected security loss.
The inputs should reflect the timing and basis of the security analysis. For example, collateral value should be the recoverable amount under the modeled state, not an unstressed market value if the scenario assumes a market decline. Likewise, the obligation should reflect the claims and reimbursements expected in that same state. The purpose is to avoid combining a severe obligation from one scenario with a default probability or collateral value taken from an unrelated average.
Worked Example: Claims Rise as the Insured and Collateral Deteriorate
A hypothetical large-deductible program has three economic scenarios. The ordinary scenario has probability 0.70, an obligation of $40,000, a 0.02 conditional default probability, and $36,000 of recoverable collateral. The stress scenario has probability 0.20, an obligation of $100,000, a 0.30 conditional default probability, and $60,000 of recoverable collateral. The severe scenario has probability 0.10, an obligation of $160,000, a 0.70 conditional default probability, and $48,000 of recoverable collateral. Calculate expected security shortfall, and compare it with an independence approximation.
Ordinary scenario. Shortfall on default is \(\$40{,}000-\$36{,}000=\$4{,}000\). Its expected contribution is \(0.70(0.02)(\$4{,}000)=\$56\).
Stress scenario. Shortfall on default is \(\$100{,}000-\$60{,}000=\$40{,}000\). Its contribution is \(0.20(0.30)(\$40{,}000)=\$2{,}400\).
Severe scenario. Shortfall on default is \(\$160{,}000-\$48{,}000=\$112{,}000\). Its contribution is \(0.10(0.70)(\$112{,}000)=\$7{,}840\). Expected shortfall across the scenarios is therefore \(\$56+\$2{,}400+\$7{,}840=\$10{,}296\).
For an independence approximation, first calculate the marginal default probability: \(0.70(0.02)+0.20(0.30)+0.10(0.70)=0.014+0.060+0.070=0.144\). The unconditional average potential shortfall is \(0.70(\$4{,}000)+0.20(\$40{,}000)+0.10(\$112{,}000)=\$2{,}800+\$8{,}000+\$11{,}200=\$22{,}000\). Their product is \(0.144(\$22{,}000)=\$3{,}168\), which is well below $10,296.
The scenario calculation is the appropriate one for the stated joint model: obligations grow, default becomes more likely, and collateral recovery falls in the severe state. Check: scenario probabilities total \(0.70+0.20+0.10=1.00\), and the three expected-loss contributions sum to $10,296. The independence approximation is not a second valid estimate of this joint model; it demonstrates the error from discarding its dependence.
Measuring and Managing Concentration
Concentration can be described using exposure shares. If counterparties have shares \(s_1,\ldots,s_n\) of total exposure, where the shares sum to 1, the Herfindahl index is \(\sum_j s_j^2\). A larger value indicates that exposure is more concentrated. Its reciprocal is sometimes interpreted as an effective number of equally sized counterparties, but it is a descriptive measure, not a loss estimate. It does not account for different default probabilities, collateral quality, or correlations.
A concentration review should also look beyond the legal counterparty. Several nominally separate insureds may rely on the same parent, operate in the same region, or be exposed to the same catastrophe or supply-chain disruption. Security can be concentrated too: a trust holding assets from one issuer, or collateral whose value is tied to the insured’s own financial condition, may lose value just when reimbursement capacity weakens.
| Concentration source | Question to ask | Why it matters |
|---|---|---|
| Counterparty | How much unpaid obligation depends on one insured or related group? | One failure can create a large recovery gap. |
| Common exposure driver | Could multiple insureds be affected by the same event or economic condition? | Defaults that appear separate may occur together. |
| Collateral issuer or asset | Could several security instruments lose value for the same reason? | Nominal diversification may not diversify recovery risk. |
| Obligation and default dependence | Do claim obligations rise in the states where default is more likely? | Average exposure and average default assumptions can understate joint loss. |
Building a Defensible Stress View
Scenario analysis is most useful when assumptions are explicit and internally consistent. A structured review can proceed as follows.
Group exposure by insured and related entities, collateral source, and material common drivers. Measure shares where appropriate, but do not treat a concentration index as a substitute for loss analysis.
Describe ordinary and adverse conditions that can affect claim obligations, insured liquidity, conditional default probability, and collateral recovery at the same time.
For each scenario, use its obligation and recoverable collateral to calculate \(\max(O-C,0)\), then apply the scenario probability and conditional default probability.
Expected loss summarizes an average across scenarios. Also inspect the size and probability of severe aggregate shortfalls, especially where counterparties or collateral share a common driver.
A common exam trap is to apply a single default probability to an average collateral amount and an average obligation even when the problem gives conditional assumptions by state. Use the conditional inputs as stated. Another trap is to conclude that splitting exposure automatically reduces expected loss. It may reduce the chance of an extreme total under independence, but only the joint distribution supports that conclusion; common-factor dependence can remove much of the benefit.
These calculations support a disciplined view of large-deductible security risk: quantify exposure, preserve the relationship among claims, credit quality, and collateral value, and distinguish average loss from the severity of adverse aggregate outcomes. The resulting analysis is aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus.
Check Your Understanding
For each item, calculate the requested measure using the stated assumptions.
- A single hypothetical insured owes $80,000, has $50,000 of recoverable collateral, and has a 0.15 default probability. Find expected shortfall. (Answer: \(\$4{,}500\).)
- Two separate insureds each owe $40,000 and each has $25,000 of recoverable collateral. Each defaults with probability 0.20, independently. Find combined expected shortfall and the probability total shortfall is $30,000. (Answer: Combined expected shortfall is \(\$6{,}000\); probability of a \(\$30{,}000\) shortfall is \(0.20^2=0.0400\).)
- A normal scenario has probability 0.75, default probability 0.04, and potential shortfall of $10,000. A stress scenario has probability 0.25, default probability 0.40, and potential shortfall of $70,000. Find joint expected shortfall, then the estimate from multiplying marginal default probability by unconditional average potential shortfall. (Answer: Joint expected shortfall is \(0.75(0.04)(\$10{,}000)+0.25(0.40)(\$70{,}000)=\$7{,}300\); the independence estimate is \([0.75(0.04)+0.25(0.40)][0.75(\$10{,}000)+0.25(\$70{,}000)]=0.13(\$25{,}000)=\$3{,}250\).)
- Three counterparties have exposure shares of 50%, 30%, and 20%. Calculate the Herfindahl index and its reciprocal effective-counterparty measure. (Answer: \(0.50^2+0.30^2+0.20^2=0.38\); reciprocal \(1/0.38\approx2.63\) equally sized counterparties.)
- A hypothetical account has an obligation of $120,000. In a stress state, the conditional default probability is 0.25 and recoverable collateral is $72,000. In a normal state, the default probability is 0.03 and collateral is $108,000. Stress has probability 0.10 and normal has probability 0.90. Calculate expected shortfall. (Answer: \(0.10(0.25)(\$48{,}000)+0.90(0.03)(\$12{,}000)=\$1{,}200+\$324=\$1{,}524\).)