Exposure Rating With Expected Loss Ratios

From Subject Premium to Expected Layer Loss

Exposure rating estimates losses for a layer by starting with the expected losses on the subject business and assigning the appropriate share of those losses to the layer. When the available starting point is subject premium and an expected loss ratio, the first step is to estimate ground-up expected losses. The second is to use the severity curve to identify what proportion of those losses falls between the layer’s attachment and its limit.

The key distinction is that a cumulative curve factor at a limit includes losses below that limit. It is not, by itself, the proportion of losses in an excess layer. A layer indication uses the difference between the factors at its upper limit and attachment.

Definition: Let $P_{\mathrm{sub}}$ be subject premium on the selected premium basis, and let $r_{\mathrm{EL}}$ be the expected loss ratio on a consistent basis. The estimated ground-up expected losses are $G=P_{\mathrm{sub}}r_{\mathrm{EL}}$. Define the cumulative expected loss proportion at limit $L$ as $C(L)=E[\min(X,L)]/E[X]$, where $X$ is a nonnegative ground-up loss severity with finite, positive mean. For a layer attaching at $a$ with limit $u$, the expected loss proportion is $C(a+u)-C(a)$, and its estimated expected losses are $G[C(a+u)-C(a)]$.

The curve proportions must describe the same loss measure as the expected loss ratio: for example, both should include or exclude allocated loss adjustment expenses in the same way. The premium, expected loss ratio, and curve must also refer to compatible business, coverage, and rating conditions. A supplied curve represents the assumed severity distribution and exposure mix; no particular parametric severity distribution is required here. These examples do not involve interest or mortality assumptions.

Why a Layer Uses a Difference of Curve Factors

For one ground-up loss $X=x$, the payment from a layer attaching at $a$ with limit $u$ is $\min((x-a)_+,u)$, where $(z)_+=\max(z,0)$. This payment can also be written as the difference between two limited-loss amounts:

\min((x-a)_+,u)=\min(x,a+u)-\min(x,a).

To verify the identity, consider the possible values of $x$. If $x\leq a$, both limited-loss amounts on the right equal $x$, so the difference is zero. If $a<x<a+u$, the difference is $x-a$. If $x\geq a+u$, the difference is $u$. These are exactly the layer payments in each case. Taking expectations and dividing by $E[X]$ gives the layer’s expected loss proportion:

\frac{E[\min((X-a)_+,u)]}{E[X]} = \frac{E[\min(X,a+u)]}{E[X]} - \frac{E[\min(X,a)]}{E[X]} = C(a+u)-C(a).

Multiplying that proportion by ground-up expected losses estimates the expected losses for the layer. This step uses expected values, so it does not require independence among claims. It does require the curve to represent the severity distribution applicable to the subject business.

A Practical Calculation Sequence

Check the bases.
Confirm that subject premium, expected loss ratio, and curve refer to the same business and consistent loss definitions.

Calculate ground-up expected losses.
Multiply subject premium by the expected loss ratio, expressed as a decimal.

Identify both curve points.
For an $a$ attaching excess of $u$ layer, use the factors at $a$ and $a+u$.

Calculate and apply the layer proportion.
Subtract the attachment factor from the upper-limit factor, then multiply the result by ground-up expected losses.

Check the result.
The layer proportion should be nonnegative and no greater than one. For adjacent layers, their proportions should add to the cumulative proportion across the combined band.

Worked Applications

Worked Example: One Layer From a Subject Expected Loss Ratio

A hypothetical insurer has subject premium of $2,400,000 and selects an expected loss ratio of 62%. Its supplied cumulative severity curve gives $C(\$1{,}000{,}000)=0.72$ and $C(\$3{,}000{,}000)=0.91$. Estimate expected losses for a $1,000,000 attaching excess of $2,000,000 layer. The curve factors are supplied directly; no parametric severity distribution is assumed.

Method. First estimate ground-up expected losses, then apply the difference between the curve factor at the layer’s upper limit and the factor at its attachment.

The ground-up expected losses are:

G=\$2{,}400{,}000(0.62)=\$1{,}488{,}000.

The layer attaches at $1,000,000 and has a $2,000,000 limit, so its upper limit is $1,000,000 + $2,000,000 = $3,000,000. Its expected loss proportion is:

C(\$3{,}000{,}000)-C(\$1{,}000{,}000)=0.91-0.72=0.19.

The expected layer losses are $ \$1{,}488{,}000(0.19)=\$282{,}720$. As a check, the expected limited losses below $3,000,000 are $ \$1{,}488{,}000(0.91)=\$1{,}354{,}080$, while those below $1,000,000 are $ \$1{,}488{,}000(0.72)=\$1{,}071{,}360$. Their difference is $282,720.

Common mistake avoided: Multiplying ground-up expected losses by 0.91 would give $1,354,080, which includes expected losses below the layer’s attachment. Subtracting the attachment factor isolates the layer. No rounding is needed.

Worked Example: Combining Class-Specific Indications

A hypothetical portfolio has two classes of business. Class A has subject premium of $1,500,000, an expected loss ratio of 60%, and curve factors $C_A(\$500{,}000)=0.50$ and $C_A(\$2{,}500{,}000)=0.88$. Class B has subject premium of $2,400,000, an expected loss ratio of 45%, and factors $C_B(\$500{,}000)=0.44$ and $C_B(\$2{,}500{,}000)=0.79$. Estimate total expected losses for a $500,000 attaching excess of $2,000,000 layer. The supplied class curves, rather than a parametric distribution, describe the respective severity assumptions.

Method. Calculate ground-up expected losses and the layer proportion separately for each class. Then add the two class-level layer indications.

For Class A, ground-up expected losses are $ \$1{,}500{,}000(0.60)=\$900{,}000$. The upper limit is $500,000 + $2,000,000 = $2,500,000, so the layer proportion is $0.88-0.50=0.38$. Class A’s expected layer losses are $ \$900{,}000(0.38)=\$342{,}000$.

For Class B, ground-up expected losses are $ \$2{,}400{,}000(0.45)=\$1{,}080{,}000$. Its layer proportion is $0.79-0.44=0.35$, so expected layer losses are $ \$1{,}080{,}000(0.35)=\$378{,}000$.

Total ground-up expected losses are $ \$900{,}000+\$1{,}080{,}000=\$1{,}980{,}000$. Total expected layer losses are $ \$342{,}000+\$378{,}000=\$720{,}000$. As a check, the combined loss-weighted layer proportion is $ \$720{,}000/\$1{,}980{,}000=0.3636$ to four decimal places; multiplying $ \$1{,}980{,}000(0.3636)$ gives $719,928, with the small difference due to rounding. Using the unrounded ratio $720{,}000/1{,}980{,}000$ gives exactly $720,000.

Common mistake avoided: Do not average the class proportions and apply that average to total expected losses unless the weights are appropriate. Here the classes have different ground-up expected losses and different curve proportions, so separate calculations preserve the correct weighting. No rounding is used in the total.

Worked Example: Checking Adjacent Layers Against the Curve

A hypothetical program has subject premium of $5,000,000 and an expected loss ratio of 54%. The supplied cumulative curve gives $C(\$250{,}000)=0.34$, $C(\$1{,}250{,}000)=0.70$, and $C(\$4{,}250{,}000)=0.93$. Estimate expected losses for two adjacent layers: $250,000 attaching excess of $1,000,000, and $1,250,000 attaching excess of $3,000,000. The curve factors summarize the assumed severity distribution; no separate distribution parameters are supplied or needed.

Method. Calculate ground-up expected losses, apply the appropriate curve difference to each layer, and check that the two proportions add to the curve difference across the combined band.

Ground-up expected losses are $ \$5{,}000{,}000(0.54)=\$2{,}700{,}000$. The first layer runs from $250,000 to $1,250,000, so its proportion is $0.70-0.34=0.36$, and its expected losses are $ \$2{,}700{,}000(0.36)=\$972{,}000$.

The second layer runs from $1,250,000 to $4,250,000. Its proportion is $0.93-0.70=0.23$, and its expected losses are $ \$2{,}700{,}000(0.23)=\$621{,}000$. The combined expected losses are $ \$972{,}000+\$621{,}000=\$1{,}593{,}000$.

For the check, the combined band’s curve proportion is $0.93-0.34=0.59$, and $ \$2{,}700{,}000(0.59)=\$1{,}593{,}000$, matching the sum. The adjacent layers join at $1,250,000, so there is neither a gap nor an overlap. No rounding is needed.

Common mistake avoided: Using $C(\$4{,}250{,}000)$ alone for the combined band would include losses below $250,000. The combined band starts at that attachment, so its factor is $0.93-0.34$, not 0.93.

Interpretation, Limitations, and Exam Traps

An expected loss ratio translates premium into an expected loss amount; it does not directly give a layer’s expected loss ratio or layer price. The severity curve allocates ground-up expected losses across ranges of loss sizes. The resulting layer loss indication is an expected loss estimate, not automatically a premium indication: expenses, commissions, risk provisions, and other pricing adjustments are outside this calculation.

The method is only as appropriate as its inputs. If the expected loss ratio refers to a different coverage, business mix, or loss definition than the curve, the multiplication and allocation may not estimate the intended layer. Likewise, a curve based on limited or censored losses should not be treated as a ground-up curve without a suitable correction. Use supplied factors on their stated basis and do not silently substitute a different basis.

Using the upper-limit factor alone. $C(a+u)$ includes losses below the attachment; subtract $C(a)$.
Applying the expected loss ratio to the layer factor first. The consistent sequence is premium times expected loss ratio, then times the layer proportion.
Using the limit as the upper loss amount. A layer attaching at $a$ with limit $u$ ends at $a+u$, not at $u$.
Combining unlike classes prematurely. When classes have different expected loss ratios or curves, calculate their layer losses separately before adding them.

Key takeaway: Convert subject premium to ground-up expected losses with the expected loss ratio. Then estimate the expected loss in an $a$ attaching excess of $u$ layer as $P_{\mathrm{sub}}r_{\mathrm{EL}}[C(a+u)-C(a)]$, provided the premium, loss, and curve bases are consistent.

Check Your Understanding

For each problem, use the stated cumulative factors as supplied; no parametric severity distribution is assumed.

A hypothetical book has subject premium of $800,000 and an expected loss ratio of 55%. The curve gives $C(\$200{,}000)=0.30$ and $C(\$1{,}000{,}000)=0.68$. Find expected losses for a $200,000 attaching excess of $800,000. (Answer: $167,200)
Subject premium is $3,200,000 and the expected loss ratio is 50%. For a $1,000,000 attaching excess of $4,000,000 layer, the curve factors are 0.60 at $1,000,000 and 0.94 at $5,000,000. Find expected layer losses. (Answer: $544,000)
Class A has premium of $1,000,000, an expected loss ratio of 70%, and a layer proportion of 0.25. Class B has premium of $2,000,000, an expected loss ratio of 40%, and a layer proportion of 0.28. Find combined expected layer losses. (Answer: $399,000)
Ground-up expected losses are $1,200,000. A curve gives factors 0.40 at $500,000, 0.82 at $2,000,000, and 0.95 at $5,000,000. Find expected losses for the $500,000 attaching excess of $1,500,000 layer and the $2,000,000 attaching excess of $3,000,000 layer, and check their sum. (Answer: $504,000 and $156,000; sum $660,000)
A hypothetical portfolio has an expected loss ratio of 50%. A layer’s curve proportion is 0.24, and its estimated expected layer losses are $360,000. Find the portfolio’s ground-up expected losses and the subject premium consistent with these inputs. (Answer: ground-up expected losses $1,500,000; subject premium $3,000,000)

What You'll Learn