Why Collectibility Matters
A treaty calculation estimates the recovery to which an insurer may be entitled under the contract. That amount is not necessarily the amount the insurer will ultimately collect. A reinsurer may be unable to pay, the parties may disagree about whether particular losses are covered, or collateral may secure only part of the obligation. The estimated recovery therefore has two separate questions behind it: what is contractually recoverable, and how much of that amount is expected to be collected?
This distinction matters when interpreting ceded and net estimates. A lower collectible amount does not reduce the insurer’s gross claim obligation to its policyholders. It changes the estimated value of the recovery from the reinsurer and, consequently, the estimated net cost after reinsurance.
The estimate \(R\) should be based on the treaty’s terms and the claim information, including any relevant limits, retentions, and definitions. The earlier material on reinsurance recoveries and changing treaty terms addresses how to estimate the contractual recovery. This tutorial focuses on uncertainty about collection after that contractual estimate has been made.
Separating Credit Risk from Coverage Disputes
Credit risk concerns whether a reinsurer can pay an amount it owes. Coverage risk concerns whether the reinsurer owes the amount under the treaty, including how a disputed interpretation or claim fact will be resolved. These risks may interact: a reinsurer’s financial condition may change while a dispute is being resolved, for example. They should not automatically be treated as independent.
A transparent estimate identifies the possible outcomes, the probability assigned to each outcome, and the amount collected in that outcome. If there are mutually exclusive outcomes \(j\), with probabilities \(p_j\) and corresponding collection amounts \(c_j\), then the expected collectible amount is:
This is an expected-value model. It requires the outcome probabilities and amounts to represent the estimate being made. It does not, by itself, describe the spread of possible outcomes or the amount needed for a separate risk measure. If credit and coverage probabilities are multiplied, the calculation must use conditional probabilities or a defensible independence assumption.
For a simplified recoverable of \(R\), suppose there is probability \(q\) of default, the reinsurer pays the full amount if it does not default, and the fraction \(r\) is collected in default. The model assumes these are the only two credit outcomes and that the stated payment fractions apply to this recoverable. Then:
The first term represents full payment without default; the second represents partial collection in default. The expected shortfall is \(R-E[C]=q(1-r)R\). This result is not appropriate if, for example, the recoverable is disputed, default outcomes differ across portions of the balance, or collateral changes the amount exposed to default. Those features should be modeled explicitly.
Worked Example: Credit Risk on an Undisputed Recoverable
A hypothetical insurer estimates a contractual recoverable of \(R=\$2{,}000{,}000\). For this example, assume the amount is undisputed, there is an 8% probability of reinsurer default, and 25% of the recoverable is collected if default occurs. If no default occurs, the full recoverable is collected. The gross ultimate claims estimate is \(G=\$5{,}400{,}000\). Estimate expected collections, the expected shortfall, and net ultimate claims.
Method: two-outcome expected collection. The probability of no default is \(1-0.08=0.92\). The probability of default is \(0.08\). The collection in the no-default outcome is \(\$2{,}000{,}000\), and the collection in default is \(0.25(\$2{,}000{,}000)=\$500{,}000\).
The expected collection from the no-default outcome is \(0.92(\$2{,}000{,}000)=\$1{,}840{,}000\). The expected collection from the default outcome is \(0.08(\$500{,}000)=\$40{,}000\). Therefore:
The expected shortfall relative to the contractual recoverable is \(\$2{,}000{,}000-\$1{,}880{,}000=\$120{,}000\). The net ultimate claim estimate is \(\$5{,}400{,}000-\$1{,}880{,}000=\$3{,}520{,}000\).
Check. The shortfall also equals \(0.08(1-0.25)(\$2{,}000{,}000)=\$120{,}000\), matching the outcome calculation. The common mistake this method avoids is subtracting the full \(\$2{,}000{,}000\) contractual recovery from gross claims even though the assumptions imply an expected collection of only \(\$1{,}880{,}000\).
Modeling Disputed Recoverables
A dispute can affect whether a recovery is owed, how much is owed, or when payment will occur. Do not treat the entire disputed amount as either certain to be paid or certain to be lost unless that is the explicit estimate. Instead, specify the potential coverage outcomes and the collection assumption conditional on each outcome.
For a disputed amount \(D\), suppose it is upheld with probability \(h\). If upheld, assume the reinsurer pays the amount with probability \(s\); if not paid, assume no recovery from that amount. Under those stated assumptions, the expected collection on the disputed portion is \(Dhs\). This multiplication is justified because \(s\) is defined as the probability of collection conditional on the dispute being upheld. If payment could be partial, or credit conditions differ by outcome, list the resulting outcomes and their collection amounts instead.
Worked Example: Combining an Uncontested Balance and a Coverage Dispute
A hypothetical insurer has a \(\$2{,}000{,}000\) contractual recoverable: \(\$800{,}000\) is undisputed and \(\$1{,}200{,}000\) is disputed. Assume the reinsurer pays an undisputed amount with probability 90%. The disputed amount is upheld with probability 75%; conditional on being upheld, it is collected with probability 90%, and otherwise no part of it is collected. The gross ultimate claims estimate is \(\$4{,}700{,}000\). Estimate expected collections and net ultimate claims.
Method: separate the portions and apply conditional probabilities. Expected collection on the undisputed portion is \(0.90(\$800{,}000)=\$720{,}000\).
For the disputed portion, the probability of both an upheld dispute and collection is \(0.75(0.90)=0.675\). Expected collection on that portion is \(0.675(\$1{,}200{,}000)=\$810{,}000\). Thus:
The expected shortfall from the \(\$2{,}000{,}000\) contractual estimate is \(\$470{,}000\). Net ultimate claims are \(\$4{,}700{,}000-\$1{,}530{,}000=\$3{,}170{,}000\).
Check. The disputed portion has expected collection of 67.5% of \(\$1{,}200{,}000\), or \(\$810{,}000\); adding the undisputed portion’s \(\$720{,}000\) gives \(\$1{,}530{,}000\). A common mistake is to apply the 90% collection probability to the disputed amount without first allowing for the 75% chance that the dispute is upheld.
Collateral and Collection Experience
Collateral can reduce the amount exposed to reinsurer credit risk, but its value depends on whether it is available, enforceable, and sufficient for the obligation. A useful estimate distinguishes the portion supported by collateral from the unsecured portion. If collateral is subject to a haircut, access delay, or a competing claim, reflect that in its expected realizable amount rather than treating its face amount as certain cash.
The simple calculation below assumes collateral is fully accessible and collectible, covers a specified part of the recoverable, and has no separate cost or delay adjustment. The unsecured portion follows a two-outcome default model. These are assumptions for the model, not general conclusions about collateral.
Worked Example: Collateral and Exposure to Default
A hypothetical insurer estimates a \(\$3{,}000{,}000\) contractual recoverable. Of this amount, \(\$1{,}100{,}000\) is supported by collateral assumed fully accessible and collectible. The remaining \(\$1{,}900{,}000\) is unsecured. Assume a 6% default probability on the unsecured amount and 20% collection of that unsecured amount in default; otherwise it is collected in full. Gross ultimate claims are estimated at \(\$7{,}600{,}000\). Estimate expected collections and net ultimate claims.
Method: separate secured and unsecured amounts. The collateral-supported amount contributes \(\$1{,}100{,}000\) to expected collections under the stated assumption. The probability of no default on the unsecured amount is \(1-0.06=0.94\). Its expected collection is:
Total expected collections are \(\$1{,}100{,}000+\$1{,}808{,}800=\$2{,}908{,}800\). The expected shortfall is \(\$3{,}000{,}000-\$2{,}908{,}800=\$91{,}200\). Net ultimate claims are \(\$7{,}600{,}000-\$2{,}908{,}800=\$4{,}691{,}200\).
Check. Only the unsecured amount contributes a shortfall: \(0.06(1-0.20)(\$1{,}900{,}000)=\$91{,}200\). Treating the entire recoverable as unsecured would ignore the assumed collateral; treating the entire recoverable as collateralized would overstate protection.
Collection experience is useful evidence, but it must be interpreted carefully. A simple historical collection ratio can be calculated for a sufficiently mature group of recoverables as amounts ultimately collected divided by amounts due for that group. The denominator, valuation date, and treatment of amounts still outstanding should be explicit. A recent low paid-to-due ratio may reflect slow settlement rather than permanent loss; a ratio based only on fully settled balances may omit difficult, unresolved disputes. Experience should therefore be segmented where relevant by reinsurer, treaty, type of dispute, age of balance, and collateral status, and interpreted alongside current facts.
For example, if \(\$1{,}000{,}000\) of a mature cohort was due and \(\$960{,}000\) was ultimately collected, the observed collection ratio is \(0.96\), or 96%. Applying that ratio mechanically to every current recoverable would assume the current balances have comparable characteristics and that the historical cohort is representative. Those assumptions require support; the ratio is not, by itself, a universal adjustment factor.
Deriving the Net Estimate and Avoiding Common Traps
The net estimate follows from separating the insurer’s gross claim obligation from its expected reinsurance collection. In each outcome, let the gross claim amount be \(G\) and the recovery collected be \(C\). The insurer’s net claim cost in that outcome is \(G-C\). Taking expectations and using linearity of expectation gives:
When \(G\) is represented by a fixed gross ultimate estimate, this becomes \(G-E[C]\). The derivation explains why the collectible amount, not automatically the contractual amount \(R\), is subtracted in an expected net estimate. It also shows why credit risk does not erase the gross claim obligation: \(G\) remains the gross amount, while uncertainty about the reinsurer affects \(E[C]\).
- Do not confuse contractual entitlement with expected collection. \(R\) is based on the treaty estimate; \(E[C]\) allows for collection uncertainty.
- Do not reduce gross claims for reinsurer credit risk. The gross estimate concerns claims owed by the insurer. Collectibility changes the recovery estimate and net amount.
- Do not multiply probabilities without checking their meaning. For a coverage-and-credit calculation, use conditional probabilities or explicitly justify independence.
- Do not count collateral twice. Identify which portion of the recoverable it secures, and apply credit-risk adjustments only to the exposed amount under the model.
- Do not label every unpaid balance uncollectible. Separate expected delay from expected failure to collect; the appropriate treatment depends on the estimate and reporting basis.
Check Your Understanding
For each problem, identify the collection assumptions before calculating. Amounts are hypothetical and are not intended to describe any company or market.
- A \(\$950{,}000\) undisputed recoverable has a 4% default probability. Assume 30% is collected in default and the full amount is collected otherwise. If gross ultimate claims are \(\$2{,}400{,}000\), find expected collections, the shortfall, and net ultimate claims. (Answer: expected collections \(\$923{,}400\); shortfall \(\$26{,}600\); net ultimate claims \(\$1{,}476{,}600\))
- A \(\$1{,}100{,}000\) recoverable consists of \(\$600{,}000\) undisputed and \(\$500{,}000\) disputed. Assume the undisputed portion is collected with probability 98%; the disputed portion is upheld with probability 80%, and, conditional on being upheld, is collected with probability 90%. If gross ultimate claims are \(\$3{,}200{,}000\), find expected collections and net ultimate claims. (Answer: expected collections \(\$948{,}000\); net ultimate claims \(\$2{,}252{,}000\))
- A \(\$2{,}000{,}000\) recoverable has \(\$400{,}000\) in collateral assumed fully accessible and collectible; the remaining \(\$1{,}600{,}000\) is unsecured. For the unsecured portion, assume a 10% default probability and 25% collection in default, with full collection otherwise. If gross ultimate claims are \(\$4{,}900{,}000\), find expected collections and net ultimate claims. (Answer: expected collections \(\$1{,}880{,}000\); net ultimate claims \(\$3{,}020{,}000\))
- A mature hypothetical cohort had \(\$2{,}500{,}000\) due, of which \(\$2{,}200{,}000\) was ultimately collected. A current recoverable is \(\$800{,}000\). If, solely for this calculation, the cohort’s observed collection ratio is assumed applicable to the current balance, estimate its expected collection and shortfall. (Answer: collection ratio 88%; expected collection \(\$704{,}000\); shortfall \(\$96{,}000\))
- A \(\$1{,}500{,}000\) disputed recoverable is upheld with probability 70%. Conditional on being upheld, assume the reinsurer pays 80% of the amount with probability 90%, and nothing with probability 10%. If it is not upheld, nothing is collected. Find the expected collection and the expected shortfall from the contractual amount. (Answer: expected collection \(\$756{,}000\); expected shortfall \(\$744{,}000\))