Pricing Limited Aggregate Layers

Why Limited Aggregate Layers Need Two Thresholds

An aggregate layer protects against total losses above an attachment point, but only up to a specified limit. Its expected payment depends on the probability that aggregate loss reaches the layer and on how far into the layer it goes. A single tail probability at the attachment does not capture both effects: the aggregate loss might enter the layer without exhausting it, or it might exceed the layer’s upper endpoint.

This tutorial considers the expected insurer payment for a hypothetical layer on a nonnegative aggregate loss. It shows how to calculate that expectation from the aggregate loss distribution and how the result relates to stop-loss premiums. The calculated expectation is an expected claim payment; it does not by itself include expenses, profit, or a risk loading.

Definition: Let $S\geq 0$ be an aggregate loss, $A\geq 0$ the layer’s attachment point, and $M>0$ its limit amount. The layer payment is $$ Y=\min\bigl((S-A)_+,M\bigr), $$ where $x_+=\max(x,0)$. The layer pays nothing when $S\leq A$, pays $S-A$ when $A<S<A+M$, and pays the full limit $M$ when $S\geq A+M$. The upper endpoint $A+M$ is the exhaustion point.

The limit $M$ is the maximum amount paid by the layer, not the aggregate-loss threshold at which the layer stops paying. For example, a layer attaching at 80,000 with a limit of 100,000 exhausts at an aggregate loss of 180,000. All amounts in a particular calculation must use the same units.

Stop-Loss Transforms and the Layer Identity

For a nonnegative aggregate loss $S$, define its stop-loss transform at retention $d\geq0$ by $\Pi_S(d)=E[(S-d)_+]$. This is the expected amount of aggregate loss above $d$. A limited layer can be represented as the difference between the excess above its attachment and the excess above its exhaustion point.

Theorem: If $E[S]<\infty$, then both stop-loss transforms below are finite and $$ E[Y]=\Pi_S(A)-\Pi_S(A+M). $$ For any nonnegative $S$, whether or not $E[S]$ is finite, the limited layer expectation is also given by $$ E[Y]=\int_A^{A+M}P(S>x)\,dx. $$

The finite-mean condition is a sufficient condition for using the difference of transforms: because $0\leq (S-d)_+\leq S$ for $d\geq0$, both transforms are finite when $E[S]$ is finite. The integral formula is more general. Its integration interval is finite, its integrand is at most one, and therefore it gives a finite expected payment even if $E[S]$ is infinite.

Why the Formulas Hold

The transform identity follows from an equality for each possible value $s$ of $S$. If $s\leq A$, both excess terms are zero. If $A<s<A+M$, their difference is $s-A$. If $s\geq A+M$, their difference is $M$. These are exactly the three possible layer payments, so

\min\bigl((s-A)_+,M\bigr) =(s-A)_+-(s-A-M)_+.

Setting $s=S$ and taking expectations gives $E[Y]=\Pi_S(A)-\Pi_S(A+M)$, provided those expectations are finite. For the tail-integral formula, at a fixed $s$, the quantity $\min((s-A)_+,M)$ is the length of the part of the interval from $A$ to $A+M$ that lies below $s$. Equivalently,

\min\bigl((s-A)_+,M\bigr) =\int_A^{A+M}\mathbf{1}_{\{s>x\}}\,dx.

Substituting $S$ for $s$, taking expectations, and interchanging expectation and integration for this nonnegative integrand gives

E[Y] =\int_A^{A+M}E[\mathbf{1}_{\{S>x\}}]\,dx =\int_A^{A+M}P(S>x)\,dx.

This derivation also explains why a finite layer payment has a finite expectation even when the aggregate loss has an infinite mean: $Y$ is bounded above by $M$.

Calculating from an Aggregate Distribution

When the aggregate loss has a discrete distribution, a direct calculation can be a useful check on the transform difference. Calculate the layer payment at each possible aggregate loss, multiply by its probability, and sum. The following example compares all three calculations: direct payments, stop-loss transforms, and the tail integral.

Worked Example: Discrete Aggregate Loss and a Limited Layer

A hypothetical portfolio has aggregate loss $S$, measured in thousands of dollars, with the following distribution: $S=0,50,120,250$ with probabilities $0.20,0.30,0.30,0.20$, respectively. A layer attaches at $A=80$ thousand dollars and has limit $M=100$ thousand dollars. Calculate its expected payment. The aggregate loss is bounded, so $E[S]<\infty$.

Method: calculate payments by outcome, then verify using stop-loss transforms. The exhaustion point is $A+M=80+100=180$ thousand dollars. The four layer payments are

\begin{aligned} S=0 &: \min((0-80)_+,100)=0,\\ S=50 &: \min((50-80)_+,100)=0,\\ S=120 &: \min((120-80)_+,100)=40,\\ S=250 &: \min((250-80)_+,100)=100. \end{aligned}

The expected payment, in thousands of dollars, is therefore

$$ 0.20(0)+0.30(0)+0.30(40)+0.20(100) =12+20=32. $$

For a transform check, $\Pi_S(80)=0.30(120-80)+0.20(250-80)=12+34=46$ thousand dollars. Also, $\Pi_S(180)=0.20(250-180)=14$ thousand dollars. Thus

\Pi_S(80)-\Pi_S(180)=46-14=32

thousand dollars, matching the direct calculation. Common mistake avoided: the second transform uses the exhaustion point $A+M=180$, not the limit amount $M=100$. In fact, $\Pi_S(100)=0.30(20)+0.20(150)=36$, so incorrectly using $\Pi_S(80)-\Pi_S(100)$ would give 10 thousand dollars, not the layer expectation.

Calculating from Aggregate Tails

If aggregate-tail probabilities are supplied over intervals, the layer expectation is the area under the tail curve between the attachment and exhaustion points. On each interval where the tail probability is constant, the area is the interval width multiplied by that probability. If a boundary falls inside an interval, split the calculation at that boundary rather than using a full interval.

Worked Example: Integrating a Piecewise Aggregate Tail

A hypothetical aggregate loss is measured in thousands of dollars. Its tail probabilities are $P(S>x)=0.75$ for $0\leq x<25$, $0.50$ for $25\leq x<75$, $0.20$ for $75\leq x<125$, and zero for $x\geq125$. A layer attaches at 40 thousand dollars and has limit 60 thousand dollars. Calculate its expected payment, and check the result using the transforms.

Method: integrate the tail only across the layer’s interval. The exhaustion point is $40+60=100$ thousand dollars. The interval from 40 to 100 crosses the tail boundary at 75. Thus the tail probability is 0.50 from 40 to 75 and 0.20 from 75 to 100. The expected payment is

$$ (75-40)(0.50)+(100-75)(0.20) =35(0.50)+25(0.20) =17.5+5=22.5 $$

thousand dollars. For the transform check, $\Pi_S(40)=35(0.50)+50(0.20)=17.5+10=27.5$ thousand dollars. At the exhaustion point, $\Pi_S(100)=25(0.20)=5$ thousand dollars. The difference is $27.5-5=22.5$ thousand dollars, as required.

This tail is consistent with aggregate-loss probabilities $P(S=0)=0.25$, $P(S=25)=0.25$, $P(S=75)=0.30$, and $P(S=125)=0.20$. A direct check gives layer payments $0,0,35,60$, respectively, for a mean of $0.30(35)+0.20(60)=10.5+12=22.5$ thousand dollars. Common mistake avoided: integrating the tail from the attachment all the way to the largest possible loss would calculate the uncapped excess above 40, not the limited layer.

A Layer from a Simple Frequency Model

The aggregate distribution need not be provided as a table. It may be obtained from a frequency model and claim amounts, after which the layer can be priced from the resulting aggregate outcomes. This calculation uses a small model so that each aggregate probability and payment can be checked explicitly.

Worked Example: A Layer on a Binomial Claim Count

A hypothetical one-period portfolio has claim count $N\sim\operatorname{Binomial}(2,0.5)$. Each claim is exactly 60 thousand dollars, so $S=60N$ thousand dollars. A layer attaches at 50 thousand dollars and has limit 70 thousand dollars. Calculate the expected layer payment. The claim count has only two possible claim opportunities, and the aggregate loss is bounded.

Method: find the aggregate distribution, then apply the layer definition. The binomial probabilities are $P(N=0)=0.25$, $P(N=1)=2(0.5)(0.5)=0.50$, and $P(N=2)=0.25$. The corresponding aggregate losses are 0, 60, and 120 thousand dollars. The exhaustion point is $50+70=120$ thousand dollars, so the layer payments are 0, 10, and 70 thousand dollars.

The expected payment is

$$ 0.25(0)+0.50(10)+0.25(70) =5+17.5=22.5 $$

thousand dollars. As a transform check, $\Pi_S(50)=0.50(60-50)+0.25(120-50)=5+17.5=22.5$ thousand dollars, while $\Pi_S(120)=0$. Their difference is 22.5 thousand dollars. The outcome at 120 exhausts the layer, so its payment is 70 rather than the excess of 70 plus any additional amount. Common mistake avoided: do not pay the entire aggregate loss above the attachment; the layer payment is capped at its stated limit.

Monotonicity and Practical Checks

For a fixed aggregate-loss distribution and fixed limit, increasing the attachment cannot increase the expected payment. To see this, fix an aggregate outcome $s$. If $A_2>A_1$, then $(s-A_2)_+\leq(s-A_1)_+$, and therefore

\min((s-A_2)_+,M)\leq\min((s-A_1)_+,M).

Taking expectations preserves this inequality. The expected payment may decrease or remain unchanged, for example if the aggregate loss never reaches either layer, but it cannot increase. Conversely, with attachment fixed, increasing the limit cannot decrease the payment for any outcome, and hence cannot decrease its expectation.

When checking a calculation, confirm that the interval of integration starts at $A$ and ends at $A+M$; that every tail probability is multiplied by the width of the interval where it applies; and that the final expected payment lies between zero and $M$. If the aggregate loss is bounded or has finite mean, also compare the result with $\Pi_S(A)-\Pi_S(A+M)$. These checks often reveal a misplaced endpoint, an inconsistent unit, or an uncapped calculation.

Key takeaway: The expected payment for a limited aggregate layer is the tail area between its attachment and exhaustion points. When the relevant stop-loss transforms are finite, the same quantity is their difference, $\Pi_S(A)-\Pi_S(A+M)$. Keep the limit amount and the upper aggregate-loss threshold distinct.

Check Your Understanding

For each problem, identify the layer’s exhaustion point before calculating. All monetary amounts are stated in thousands of dollars unless otherwise noted.

A hypothetical aggregate loss takes values 0, 40, 90, and 160 with probabilities 0.10, 0.20, 0.40, and 0.30. A layer attaches at 50 and has limit 70. Calculate its expected payment by outcome. (Answer: 37 thousand dollars)
A hypothetical aggregate tail is $P(S>x)=0.80$ for $0\leq x<30$, $0.50$ for $30\leq x<80$, $0.10$ for $80\leq x<140$, and zero for $x\geq140$. A layer attaches at 60 and has limit 50. Calculate its expected payment by integrating the tail over the layer interval. (Answer: 13 thousand dollars)
A hypothetical aggregate loss equals 0, 100, or 300 with probabilities 0.40, 0.35, or 0.25. A layer attaches at 80 and has limit 150. Calculate the expected payment and verify it is no greater than the limit. (Answer: 44.5 thousand dollars)
A hypothetical portfolio has $N\sim\operatorname{Binomial}(2,0.5)$, with each claim exactly 40 thousand dollars. A layer attaches at 30 and has limit 50. Find the aggregate outcomes and probabilities, then calculate the expected layer payment. (Answer: 17.5 thousand dollars)
A nonnegative hypothetical aggregate loss has $P(S>x)=1$ for $0\leq x<1$ and $P(S>x)=1/x$ for $x\geq1$, with amounts in dollars. Its mean is infinite. A layer attaches at $A=1$ and has limit $M=2$. Use the tail integral, not a subtraction of two infinite transforms, to find its expected payment. (Answer: $\ln(3)$ dollars, approximately $1.0986)

What You'll Learn