Counterparty Default and Recovery in ART

Why Protection Can Fall Short

An alternative risk transfer (ART) contract may promise payment after a covered loss, but the promise is not the same as cash in hand. If the counterparty defaults before paying, the amount ultimately collected depends on the claim due, collateral available, and recovery from the defaulted counterparty. The time between contract inception and claim payment matters too: a longer period can mean a greater chance of default, and collateral may be worth less or take time to liquidate.

The objective is to measure recoverable protection: the amount expected to be collected under the stated default, collateral, and recovery assumptions. This is a credit-risk estimate, not a guarantee of payment. It is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Definition: Exposure at default (EAD) is the amount owed when a counterparty defaults, before applying recovery from the counterparty's other assets. In a collateralized arrangement, the remaining unsecured exposure is the claim due less collateral value available to meet it, floored at zero. Loss given default (LGD) is the portion of that exposure not recovered after default.

A Basic Recovery Model

First consider a fixed claim amount $R$ due at one settlement date. Assume the collateral has already been adjusted for any haircut and is available at that date, with value $C$. The amount of the claim not covered by collateral is $U=\max(R-C,0)$. Let $p$ be the probability of counterparty default by the settlement date, and let $\rho$ be the fraction of the unsecured exposure recovered from the defaulted counterparty. Assume collateral is applied to the claim first, with no dispute, shortfall, or competing claim on it.

U=\max(R-C,0)

If there is no default, the full claim $R$ is paid. If default occurs, collateral contributes $\min(R,C)$, and recovery on the unsecured amount contributes $\rho U$. Thus, conditional on default, the amount collected is $\min(R,C)+\rho U$. The model treats $C$ and $\rho$ as known amounts or rates; it does not model uncertainty in collateral value or the legal recovery process.

\begin{aligned} \text{Expected collectible protection} &=(1-p)R+p\{\min(R,C)+\rho U\},\\ \text{Expected shortfall from the promised claim} &=p(1-\rho)U. \end{aligned}

The second expression follows because, when default occurs, the claim shortfall is $R-\min(R,C)-\rho U$. Since $R=\min(R,C)+U$, this shortfall is $(1-\rho)U$. When default does not occur, the shortfall is zero. Weighting the default shortfall by its probability $p$ gives the expected shortfall.

The formulas assume a single claim settlement and a default probability appropriate to that settlement date. For multiple claims, changing exposure, or dependence between claim size and default, calculate the shortfall across possible outcomes rather than automatically multiplying an average exposure by an average default probability.

Worked Example: Applying a Haircut Before Estimating Recovery

A hypothetical ART contract has a fixed claim of $10.00 million due one year from inception. By that date, the counterparty has a 4% default probability. The contract has $6.00 million of collateral subject to a 10% haircut. If the counterparty defaults, assume recovery on the remaining unsecured exposure is 25%. All amounts are measured at the same settlement date, so no discounting is needed. Treat default as a Bernoulli event with probability 0.04, and treat the claim and collateral amounts as fixed.

Method. Haircut the collateral first, then calculate unsecured exposure, default recovery, and expected collectible protection. This avoids treating the full collateral balance as cash available for the claim.

Step 1: Value the collateral after the haircut. The haircut is $0.10\times \$6.00=\$0.60$ million, so the adjusted collateral is $\$6.00-\$0.60=\$5.40$ million.

Step 2: Calculate the unsecured exposure. $U=\$10.00-\$5.40=\$4.60$ million.

Step 3: Calculate collection if default occurs. The collateral supplies $5.40 million and recovery on the unsecured amount is $0.25\times \$4.60=\$1.15$ million. Total default recovery is $\$5.40+\$1.15=\$6.55$ million.

Step 4: Calculate expected collectible protection. The probability of no default is $1-0.04=0.96$. Therefore, expected collection is $0.96\times \$10.00+0.04\times \$6.55=\$9.862$ million.

Check and common mistake. The expected shortfall is $0.04\times(1-0.25)\times \$4.60=\$0.138$ million, and $\$10.00-\$0.138=\$9.862$ million, matching the calculation. A wrong calculation would use $6.00 million as the collateral value, producing unsecured exposure of $4.00 million and overstating protection. The 10% haircut must be applied before calculating the unsecured amount.

Claim Timing and Default Probability

The probability of default should match the period during which the counterparty must remain able to pay. One standard simplifying model assumes a constant default hazard $\lambda$ per year: conditional on survival to any time, the probability of default in a short interval of length $\Delta t$ is approximately $\lambda\Delta t$. Under this assumption, the time to default has an exponential distribution, and the probability of default by time $t$ is $1-e^{-\lambda t}$.

P(\text{default by }t)=1-e^{-\lambda t}

This cumulative probability increases with $t$. It is not generally correct to use a one-year default probability unchanged for a claim due several years later. The constant-hazard model is a useful illustration, but it is not appropriate if the hazard changes over time unless the model is adjusted accordingly.

Collateral timing needs similar care. If collateral proceeds are delayed beyond the claim due date, they do not meet the claim in full at that date. In a simple deterministic calculation, proceeds received $\tau$ years after the claim date can be converted to their value at the claim date using the effective annual interest rate $i$, namely by dividing by $(1+i)^\tau$. This assumes the stated rate is suitable for discounting those proceeds and ignores other liquidity costs.

Worked Example: A Later Claim and Delayed Collateral

A hypothetical ART contract is expected to pay a fixed $8.00 million claim two years after inception. Assume the counterparty's time to default has a constant hazard of $\lambda=0.05$ per year, so default time is exponential. A collateral account has a nominal balance of $5.00 million, but an 8% haircut applies and liquidation proceeds arrive six months after the claim due date. Use an effective annual interest rate of $i=0.04$ to value those proceeds at the claim date. Recovery on any remaining unsecured exposure is 20%. Assume the stated collateral and recovery assumptions apply if default occurs by the claim settlement date.

Method. Calculate default probability over the full two-year period, then value delayed collateral at the claim date before calculating unsecured exposure.

Step 1: Find default probability by year two. $p=1-e^{-0.05(2)}=1-e^{-0.10}=0.0952$, to four decimal places.

Step 2: Haircut and discount the collateral proceeds. After the haircut, proceeds are $\$5.00(1-0.08)=\$4.60$ million. Their value at the claim date is $\$4.60/(1.04)^{0.5}=\$4.5107$ million, rounded to four decimal places.

Step 3: Calculate unsecured exposure and recovery if default occurs. Unsecured exposure is $\$8.00-\$4.5107=\$3.4893$ million. Recovery on that amount is $0.20\times \$3.4893=\$0.6979$ million. Total collection after default is $\$4.5107+\$0.6979=\$5.2086$ million.

Step 4: Calculate expected collection. Expected shortfall after default is $\$8.00-\$5.2086=\$2.7914$ million. Expected shortfall across the two-year period is $0.0952\times \$2.7914$, using the unrounded default probability, or approximately $0.2656 million. Expected collectible protection is $\$8.00-\$0.2656=\$7.7344$ million, rounded to four decimal places.

Check and common mistake. Default recovery plus default shortfall is $\$5.2086+\$2.7914=\$8.0000$ million, as required. Using the one-year hazard value 0.05 as the two-year default probability would understate expected shortfall. The hazard is a rate; the cumulative two-year probability is $1-e^{-0.10}$, not simply the one-year rate.

Funded and Unfunded Protection

In a funded structure, assets are set aside to support payments, often as collateral. Funding can reduce reliance on a future promise from an unsecured counterparty, but it does not make protection automatically risk-free. The collateral may be impaired by a haircut, lose value, be illiquid, be difficult to access, or be subject to legal or operational delay. Any amount not covered by realizable collateral remains exposed to counterparty recovery risk.

In an unfunded structure, payment depends more directly on the counterparty's ability and obligation to perform when a claim is due. If there is no collateral, the full claim is unsecured exposure before applying recovery. The same expected-recovery framework can compare funded and unfunded arrangements, provided each arrangement's default probability, collateral access, and recovery assumptions are specified consistently. A comparison of expected collection alone does not account for the cost of funding or other contract terms.

Worked Example: Comparing Funded and Unfunded Recoveries

A hypothetical ART cover pays $12.00 million if a specified loss event occurs. The event has probability 30%. Conditional on the event, the claim is due after one year, and counterparty default by that date has probability 6%. Treat these as a Bernoulli event and a conditional Bernoulli default event. If default occurs, recovery on unsecured exposure is 30%. Compare an unfunded arrangement with a funded arrangement holding $7.00 million of collateral subject to a 5% haircut. Assume the collateral is fully available at the claim date if default occurs, and use no discounting because all amounts are valued at that date.

Method. First compare expected collection conditional on the loss event, then multiply by the event probability to obtain unconditional expected collection. This keeps the claim-event probability separate from the conditional counterparty default probability.

Funded arrangement, step 1: Calculate adjusted collateral and unsecured exposure. Adjusted collateral is $\$7.00(1-0.05)=\$6.65$ million. Unsecured exposure is $\$12.00-\$6.65=\$5.35$ million.

Funded arrangement, step 2: Calculate collection if default occurs. Unsecured recovery is $0.30\times \$5.35=\$1.605$ million. Total default collection is $\$6.65+\$1.605=\$8.255$ million.

Funded arrangement, step 3: Calculate expected collection. Conditional on the event, expected collection is $0.94\times \$12.00+0.06\times \$8.255=\$11.7753$ million. Unconditional expected collection is $0.30\times \$11.7753=\$3.53259$ million.

Unfunded arrangement. With no collateral, the full $12.00 million is unsecured. Default collection is $0.30\times \$12.00=\$3.60$ million. Conditional expected collection is $0.94\times \$12.00+0.06\times \$3.60=\$11.496$ million. Unconditional expected collection is $0.30\times \$11.496=\$3.4488$ million.

Check and common mistake. The funded arrangement's unconditional expected collection exceeds the unfunded arrangement's by $\$3.53259-\$3.4488=\$0.08379$ million. A common error is to compare $11.7753 million with $3.4488 million: the first is conditional on an event, while the second is unconditional. Both amounts must use the same conditioning basis before comparison.

Dependence, Residual Risk, and Practical Limits

The basic formula $p(1-\rho)U$ is appropriate when $p$, $\rho$, and $U$ are fixed for the claim being analyzed, and $p$ is the relevant default probability for that claim. In practice, claim size and counterparty strength may worsen together. For example, a severe market disruption could increase the amount due under an ART contract while also weakening the counterparty. In that case, default probability and exposure are dependent, so multiplying separate averages can misstate expected shortfall. The calculation should instead consider joint outcomes and their probabilities.

The estimate also depends on what “collateral available” means. A face amount in an account is not necessarily equal to value recoverable at the claim date. Haircuts, liquidation delays, market movements, access restrictions, and competing claims may all reduce the amount that can be applied. Similarly, a recovery rate on unsecured exposure is an assumption about collection after default, not an additional guarantee. Results should state whether recovery is conditional on default and what portion of the claim it applies to.

A useful review sequence is to establish the claim amount and payment date, identify default probability through that date, value collateral available at the claim date, calculate unsecured exposure, and then apply recovery assumptions. Keeping these steps separate helps identify whether the dominant residual risk comes from counterparty credit, collateral quality, or timing.

Set the claim and date.
Identify the amount due and the date on which the counterparty must pay.

Measure default risk through that date.
Use a probability appropriate to the period and, where relevant, conditional on the claim event.

Value usable collateral.
Apply haircuts and account for timing so collateral is measured on a consistent claim-date basis.

Calculate residual exposure and recovery.
Subtract available collateral from the claim, apply the assumed recovery to the unsecured amount, and check that total collection cannot exceed the claim.

Key takeaway: Funded ART can reduce, but does not necessarily eliminate, counterparty default risk. Recoverable protection depends on the claim due, default probability at its payment date, collateral that can actually be realized, and recovery on any unsecured balance.

Check Your Understanding

For each problem, state the amount of expected collectible protection or expected shortfall requested, using the assumptions given.

A hypothetical cover has a fixed $5.00 million claim due at the valuation date. Default probability is 2%, net collateral available is $3.00 million, and unsecured recovery after default is 40%. Find expected shortfall. (Answer: $0.0240 million, or $24,000)
A $9.00 million claim is due in one year. Default probability by that date is 8%. Collateral with a $4.00 million face value has a 15% haircut, and unsecured recovery is 25%. Find expected collectible protection. (Answer: $8.6640 million)
A hypothetical event has probability 25% and produces a $20.00 million claim. Conditional on the event, default probability by payment is 3%. Net collateral is $12.00 million, and unsecured recovery is 20%. Find unconditional expected collectible protection. (Answer: $4.9520 million)
A $6.00 million claim is due in three years. Default time has a constant hazard of 0.04 per year, with an exponential distribution. There is no collateral, and recovery after default is 30%. Find expected collectible protection, using four decimal places in millions. (Answer: $5.5251 million)
A $10.00 million claim is due at a specified date. Collateral with a $6.00 million face value has a 10% haircut, and proceeds arrive one year after the claim date. Use an effective annual interest rate of 5% to value proceeds at the claim date. Default probability by the claim date is 5%, and unsecured recovery is 40%. Find expected collectible protection, in millions to four decimal places. (Answer: $9.8543 million)

What You'll Learn