Why Conditioning and Coverage Timing Matter
An aggregate payment depends on two separate ingredients: how many claims occur and how much the insurer pays on each claim. Coverage wording then determines whether a deductible or limit applies to each claim or to the total over a period. Errors arise when these ingredients are combined in the wrong order—for example, using a result conditional on a claim count as though it were unconditional, or subtracting a deductible both claim by claim and again from the total.
A reliable solution identifies the random variables, writes the contract’s payment rule, and then conditions or sums at the correct level. The examples below use hypothetical policies and claim models. The distinctions are aligned with the SOA Exam FAM (Fundamentals of Actuarial Mathematics) syllabus.
Condition on the Count, Then Average Over It
For a fixed count \(N=n\), the aggregate payment is the sum of \(n\) claim payments. If the payment amounts are independent with common mean \(E[Y]\), then its conditional expected value is \(nE[Y]\). The unconditional mean must average this conditional value over all possible counts.
The last equality uses the assumption that the claim payment distribution does not depend on the count \(N\). More generally, the law of total probability gives \(P(S\in B)=\sum_n P(S\in B\mid N=n)P(N=n)\) for any event \(B\). Omitting the weights \(P(N=n)\) leaves a conditional probability, not the required unconditional one.
Worked Example: Removing the Conditioning on a Poisson Count
A hypothetical portfolio has \(N\sim\operatorname{Poisson}(2)\) claims in a year. Each gross claim is exactly $100. The policy has an ordinary per-claim deductible of $40 and no payment limit. Find the expected aggregate insurer payment and the probability that the aggregate payment is exactly $120. Name the error avoided.
Method. First calculate the payment on one claim. The insurer pays \(Y=(X-40)_+\), where \(z_+=\max(z,0)\). Thus \(Y=(100-40)=\$60\) for every claim.
The expected claim count is \(E[N]=2\), so \(E[S]=E[N]E[Y]=2(\$60)=\$120\).
Because each claim produces exactly $60, the aggregate is exactly $120 if and only if \(N=2\). A Poisson random variable with parameter 2 has \(P(N=2)=e^{-2}2^2/2!=2e^{-2}\). Since \(e^{-2}\approx0.135335\), the probability is \(2(0.135335)=0.270670\), or 0.2707 to four decimals.
The result is checked by noting that \(N=1\) produces $60 and \(N=3\) produces $180, so no other claim count produces $120. A common error is to report \(P(S=120\mid N=2)=1\) as the answer. That is conditional on exactly two claims; the unconditional answer must multiply by \(P(N=2)\).
Apply Coverage at the Level the Contract Specifies
For an ordinary per-claim deductible \(d\) and a maximum insurer payment \(u\) on each claim, the payment on a claim of amount \(x\) is \(Y=\min((x-d)_+,u)\). The deductible is applied to each claim, and the payment cap is applied to each resulting payment. An annual aggregate deductible or annual aggregate limit is a different contract term: it acts on the total loss or total payment over the period.
| Coverage term | Level of application | Typical operation |
|---|---|---|
| Per-claim deductible \(d\) | Each claim separately | Replace \(x\) with \((x-d)_+\) |
| Per-claim payment limit \(u\) | Each claim separately | Cap each payment at \(u\) |
| Aggregate deductible \(D\) | Total over the period | Apply to the aggregate amount specified in the contract |
| Aggregate limit \(U\) | Total over the period | Cap the total insurer payment at \(U\) |
The order and level are not matters of convention to guess: they come from the stated contract. In particular, “limit” must be interpreted carefully. The formula above treats \(u\) as a maximum insurer payment per claim. If the contract instead caps the covered loss amount before the deductible, the payment rule must reflect that wording.
Worked Example: Three Claims Under Per-Claim Terms
A hypothetical policy covers three claims of $40, $90, and $160. It has a $50 deductible per claim and a $100 maximum insurer payment per claim. Find the total insurer payment. Then compare it with a different contract having a $50 annual aggregate deductible and a $100 annual aggregate payment limit. Identify the double-deductible error to avoid.
Method for the stated per-claim contract. Apply \(Y=\min((x-50)_+,100)\) separately to each claim.
For the $40 claim, \((40-50)_+=0\), so the payment is \(\min(0,100)=\$0\). For the $90 claim, \((90-50)_+=40\), so the payment is \(\min(40,100)=\$40\). For the $160 claim, \((160-50)_+=110\), so the payment is \(\min(110,100)=\$100\).
The aggregate payment under these per-claim terms is \(0+40+100=\mathbf{\$140}\). The third payment is capped at $100; the first claim receives no payment.
Under the different aggregate contract, first total the gross losses: \(40+90+160=\$290\). Applying the $50 annual aggregate deductible gives \(290-50=\$240\); applying the $100 annual aggregate payment limit gives \(\min(240,100)=\mathbf{\$100}\).
These are distinct coverage designs, so their different answers are expected. A double-deductible error would calculate the per-claim payments \(0+40+100=140\), then subtract another $50 and report $90 for the per-claim contract. That extra subtraction is not warranted: the contract specifies per-claim deductibles only. It would be appropriate to apply an additional aggregate deductible only if the contract explicitly included one.
Payment Thresholds and Probability Mass
For an ordinary deductible, a claim exactly equal to the deductible produces zero payment. This matters especially when the loss distribution has probability concentrated at particular values. At a threshold, \(P(X\ge d)\) and \(P(X>d)\) can differ substantially; the first includes claims exactly equal to \(d\), while the second does not.
If \(Y=(X-d)_+\), then \(Y=0\) whenever \(X\le d\). Consequently, \(P(Y=0)=P(X\le d)\) and \(P(Y>0)=P(X>d)\). A payment limit can also create an atom: if \(Y=\min((X-d)_+,u)\), every loss at least \(d+u\) produces payment \(u\), so \(P(Y=u)=P(X\ge d+u)\).
Worked Example: Mass at the Deductible and Payment Limit
A hypothetical claim amount \(X\) has the following distribution: \(P(X=0)=0.2\), \(P(X=500)=0.5\), and \(P(X=1{,}000)=0.3\). The policy has an ordinary deductible of $500 and a per-claim payment limit of $500. Find the payment distribution, the probability of a positive payment, the expected payment, and the 0.75 quantile of payment. Use the quantile definition \(\inf\{y:F_Y(y)\ge0.75\}\). Name the threshold error avoided.
Method. For each possible loss, use \(Y=\min((X-500)_+,500)\). If \(X=0\), then \(Y=0\). If \(X=500\), then \(Y=0\) because the loss only reaches the deductible. If \(X=1{,}000\), then the amount above the deductible is $500, so \(Y=500\).
Therefore, \(P(Y=0)=0.2+0.5=0.7\) and \(P(Y=500)=0.3\). The probability of a positive payment is \(P(Y>0)=0.3\). The expected payment is \(E[Y]=0(0.7)+500(0.3)=\$150\).
The distribution function satisfies \(F_Y(0)=0.7\) and \(F_Y(500)=1\). Since \(F_Y(0)<0.75\) but \(F_Y(500)\ge0.75\), the 0.75 quantile is $500. The results are consistent: the payment is zero with probability 0.7 and $500 with probability 0.3.
A common threshold error is to calculate the probability of payment as \(P(X\ge500)=0.8\). That includes the $500 claims, which produce no payment under an ordinary deductible. The correction is \(P(X>500)=0.3\). The equality case belongs to the zero-payment mass.
A Reliable Sequence for Integrated Calculations
When a problem combines a random count with coverage terms, separate the model from the contract before calculating. The following sequence prevents most conditioning and timing errors.
Identify the claim count distribution, the claim amount distribution, and any independence assumptions. Keep a conditional statement such as “given \(N=n\)” explicit.
Write the payment rule for one claim, or the rule for the aggregate, according to whether the deductible and limit are per claim or annual aggregate terms.
Check whether the contract pays when the loss equals a deductible or reaches a limit. Use \(>\) or \(\ge\) to match the event being calculated.
Weight conditional probabilities by the probabilities of the claim counts, or use the law of total expectation. Confirm that the resulting amount and event agree with the coverage rule.
A useful final check is to calculate a small set of claim-level payments directly before aggregating them. If an aggregate deductible or limit appears in the calculation, verify that it is actually stated as an aggregate term. For probability calculations, distinguish an event conditional on a count from the same event over the full model, and inspect any point masses at deductible or limit thresholds.
Check Your Understanding
For each problem, state the coverage level and keep strict and non-strict threshold events distinct.
- A hypothetical portfolio has \(N\sim\operatorname{Poisson}(1.5)\) claims. Each claim is exactly $80, with a $20 per-claim deductible and no limit. What is the probability that the aggregate insurer payment is exactly $120? (Answer: 0.2510)
- Three hypothetical claims are $30, $75, and $140. A policy has a $25 deductible per claim and an $80 maximum insurer payment per claim. Find the aggregate insurer payment. (Answer: $135)
- A claim amount is $100 with probability 0.4 and $200 with probability 0.6. With an ordinary $100 deductible and no limit, find the probability of a positive payment and the expected payment. (Answer: 0.6000 and $60)
- Two losses are $50 and $90, and the policy has a $40 deductible per claim with no limit. Find the correct total payment. What incorrect result would come from subtracting a second $40 as though there were also an aggregate deductible? (Answer: Correct payment $60; incorrect double-deductible result $20)
- A hypothetical portfolio has \(N\sim\operatorname{Binomial}(2,0.5)\) claims. Each claim is exactly $100, with a $30 per-claim deductible and no limit. What is the probability that the aggregate insurer payment is exactly $70? (Answer: 0.5000)