Separate Credibility Weights for Primary and Excess Losses

Why Primary and Excess Losses Can Have Separate Weights

A primary-excess experience rating plan separates losses at a selected split point. The primary component and the excess component may have different expected loss amounts, and the plan may set a different full-credibility standard for each. A single weight applied to both components would ignore those differences. Separate credibility weights allow each component’s experience to be compared with its own standard.

This tutorial uses the square-root plan-weight method. The inputs are the expected loss amount for each component over the experience period and that component’s full-credibility standard, expressed on the same basis. The weights do not determine how a claim is divided at the split point; they are applied after the component loss amounts and expected losses have been established.

Definition: For component $j$, let $E_j$ be its expected loss amount for the experience period, and let $T_j$ be its full-credibility standard, expressed as an expected loss amount on the same basis. Its square-root credibility weight is $$ Z_j=\min\left(1,\sqrt{\frac{E_j}{T_j}}\right). $$ For a primary-excess plan, calculate $Z_P$ using $E_P$ and $T_P$, and calculate $Z_E$ using $E_E$ and $T_E$. The subscripts $P$ and $E$ denote primary and excess. The weight is between 0 and 1 when $E_j$ and $T_j$ are nonnegative and $T_j>0$.

The expected loss amount $E_j$ is not the actual loss amount observed in the experience period. It is the expected amount for the relevant component, after the applicable exposure, period, and loss adjustments have been put on a consistent basis. The standard $T_j$ is the amount at which the plan assigns full credibility to that component’s experience. Both amounts must use the same currency units and the same definition of the component.

Once the weights are known, a component’s credibility-weighted loss amount can be calculated as $Z_jA_j+(1-Z_j)E_j$, where $A_j$ is the actual loss amount for that component. This blends actual experience with expected losses; it does not blend the primary actual loss with the excess expected loss. The formula assumes the actual and expected amounts are on a comparable basis.

Why the Square-Root Ratio Is Used

The square-root rule reflects the way the uncertainty in an aggregate experience measure typically declines as the volume of experience increases. For a component with a stable, comparable process, suppose the variance of its average experience measure is proportional to the reciprocal of its expected loss volume. The constant of proportionality can differ by component, but within a component the standard error then varies inversely with the square root of volume.

At full credibility, the component has volume $T_j$. At the observed expected volume $E_j$, the ratio of the full-standard error to the current standard error is

\frac{\operatorname{SE}(T_j)}{\operatorname{SE}(E_j)} = \frac{c_j/\sqrt{T_j}}{c_j/\sqrt{E_j}} = \sqrt{\frac{E_j}{T_j}},

where $c_j$ is the same component-specific constant in both standard errors. This gives the uncapped square-root weight. If the expected volume reaches or exceeds the full-credibility standard, the weight is set to 1 rather than allowed to exceed 1. This justification relies on comparable experience and the volume-based square-root rule; it does not establish that every line or component must use the same standard.

The method does not apply mechanically if $E_j$ and $T_j$ are defined on different bases, if the standard is not an expected loss amount compatible with the input, or if the plan prescribes another credibility formula. A zero expected amount gives a zero weight under this rule when $T_j>0$. A negative expected loss amount or a nonpositive standard is not a valid input for this calculation.

Calculate Each Component Independently

Identify the inputs by component.
Record $E_P,T_P$ and $E_E,T_E$ separately. Confirm that each pair has matching units, period, and component definition.

Calculate the volume ratio.
For each component, divide its expected loss amount by its own full-credibility standard: $E_j/T_j$.

Take the square root and apply the cap.
Set $Z_j=\min(1,\sqrt{E_j/T_j})$. Do not use the primary standard for excess losses or vice versa.

Use each weight with its own losses.
Combine $A_P$ with $E_P$ using $Z_P$, and $A_E$ with $E_E$ using $Z_E$, if calculating credibility-weighted amounts.

Worked Example: Different Weights Below Full Credibility

A hypothetical rating plan has expected primary losses of $720,000 and expected excess losses of $320,000. The plan’s full-credibility standards are $2,000,000 for primary and $2,000,000 for excess. Actual primary losses are $850,000 and actual excess losses are $250,000. Calculate both weights and the credibility-weighted amounts. Use the square-root plan-weight method. Assume no discounting or mortality adjustment; the hypothetical experience amounts are on consistent bases, and the plan uses the square-root rule separately for the two components.

Primary weight. Divide the primary expected amount by the primary standard:

\frac{E_P}{T_P}=\frac{720{,}000}{2{,}000{,}000}=0.36, \qquad Z_P=\sqrt{0.36}=0.60.

Excess weight. Use the excess expected amount and excess standard:

\frac{E_E}{T_E}=\frac{320{,}000}{2{,}000{,}000}=0.16, \qquad Z_E=\sqrt{0.16}=0.40.

Primary credibility-weighted amount. Apply the primary weight to actual primary losses and the remaining weight to expected primary losses:

0.60(850{,}000)+0.40(720{,}000) =510{,}000+288{,}000 =\$798{,}000.

Excess credibility-weighted amount. Apply the excess weight to actual excess losses and the remaining weight to expected excess losses:

0.40(250{,}000)+0.60(320{,}000) =100{,}000+192{,}000 =\$292{,}000.

The combined credibility-weighted amount is $\$798{,}000+\$292{,}000=\$1{,}090{,}000$. The component weights differ because the expected excess amount is a smaller share of its standard. This method avoids applying the primary weight to the excess experience simply because the two components belong to the same plan. Check: both calculated weights are between 0 and 1, and each weighted amount lies between its actual and expected component amounts.

Why the Standards Matter as Much as the Expected Amounts

A component’s weight depends on the ratio $E_j/T_j$, not on its expected amount alone. Two components with equal expected losses can receive different weights if their standards differ. Conversely, different expected amounts can produce equal weights if their ratios to their respective standards are equal. Separate standards can reflect different plan-specific credibility criteria for the components; their use should follow the stated plan assumptions, not an assumption that one component is inherently more credible.

A component at or above its standard receives full credibility under this rule. A component below its standard receives a partial weight. The cap matters because the square root of a ratio greater than 1 is also greater than 1, while a credibility weight above 1 would assign more than full weight to actual experience and a negative weight to expected experience in the blend.

Worked Example: One Component Exceeds Its Standard

A hypothetical plan has expected primary losses of $2,000,000 and a primary full-credibility standard of $500,000. Expected excess losses are $500,000, with an excess standard of $2,000,000. Actual losses are $1,700,000 primary and $700,000 excess. Calculate the weights and weighted amounts. Use the square-root plan-weight method. Assume no discounting or mortality adjustment and that all amounts are consistent hypothetical inputs. The common mistake to avoid is reversing the expected-amount-to-standard ratio.

Primary weight. The ratio is expected primary losses divided by the primary standard:

\frac{E_P}{T_P}=\frac{2{,}000{,}000}{500{,}000}=4, \qquad \sqrt{4}=2.

Since the uncapped result exceeds 1, apply the cap: $Z_P=\min(1,2)=1$. The primary component receives full credibility, so its weighted amount is

1(1{,}700{,}000)+0(2{,}000{,}000)=\$1{,}700{,}000.

Excess weight.

\frac{E_E}{T_E}=\frac{500{,}000}{2{,}000{,}000}=0.25, \qquad Z_E=\sqrt{0.25}=0.50.

The weighted excess amount is

0.50(700{,}000)+0.50(500{,}000) =350{,}000+250{,}000 =\$600{,}000.

The combined weighted amount is $\$1{,}700{,}000+\$600{,}000=\$2{,}300{,}000$. Check: primary is at full credibility; excess is halfway between its actual and expected amounts. The ratio $2{,}000{,}000/500{,}000$ is 4, not 0.25. Using its inverse would incorrectly produce a primary weight of 0.50 before capping.

Standards Can Make Weights Differ Even at Similar Volumes

The two weights are not required to match. The expected amounts and the standards both matter, and the calculation is component-specific. A plan should not infer the excess weight from the primary weight unless it explicitly specifies a relationship between the two standards and expected amounts.

Worked Example: Use the Plan’s Separate Standards

A hypothetical plan has expected primary losses of $900,000 and expected excess losses of $400,000. Its stated full-credibility standards are $1,600,000 for primary and $2,500,000 for excess. Actual losses are $1,020,000 primary and $520,000 excess. Find both weights and the combined weighted amount. Use the square-root plan-weight method. Assume no discounting or mortality adjustment, and assume the stated standards apply to the corresponding loss components on consistent hypothetical bases. The method avoids substituting a shared standard for the two plan-specific standards.

Primary weight.

\frac{E_P}{T_P}=\frac{900{,}000}{1{,}600{,}000}=0.5625, \qquad Z_P=\sqrt{0.5625}=0.75.

The primary weighted amount is

0.75(1{,}020{,}000)+0.25(900{,}000) =765{,}000+225{,}000 =\$990{,}000.

Excess weight.

\frac{E_E}{T_E}=\frac{400{,}000}{2{,}500{,}000}=0.16, \qquad Z_E=\sqrt{0.16}=0.40.

The excess weighted amount is

0.40(520{,}000)+0.60(400{,}000) =208{,}000+240{,}000 =\$448{,}000.

The combined weighted amount is $\$990{,}000+\$448{,}000=\$1{,}438{,}000$. Check: $0.75^2=0.5625$ and $0.40^2=0.16$, matching the respective expected-to-standard ratios. The weights differ because the two ratios differ, not simply because one component has a larger expected amount.

Common Mistakes and Practical Checks

Reversing the ratio. The rule uses expected amount divided by full-credibility standard. If expected losses are $2,000,000 and the standard is $500,000, the ratio is 4, not 0.25.
Forgetting the cap. If $E_j/T_j>1$, the square root exceeds 1. Set the weight to 1; do not use a weight greater than full credibility.
Using one component’s inputs for the other. Calculate primary from $E_P,T_P$ and excess from $E_E,T_E$. Do not borrow a standard merely because the components are part of one plan.
Mixing actual and expected amounts in the ratio. The square-root weight uses the expected amount and full-credibility standard. Actual losses enter only when applying the weight to form a credibility-weighted amount.
Mixing units or bases. A dollar amount cannot be compared with a standard stated in thousands without conversion. The experience period and component definition must also match.

A useful arithmetic check is to square each uncapped weight: it should equal $E_j/T_j$. For a weight capped at 1, verify that the ratio was at least 1. When $0\leq Z_j\leq1$, the weighted amount $Z_jA_j+(1-Z_j)E_j$ must lie between $A_j$ and $E_j$. These checks catch many calculation and assignment errors.

Key takeaway: Calculate primary and excess credibility weights separately using each component’s expected loss amount and its own full-credibility standard. The square-root ratio is capped at 1, and each resulting weight applies only to the matching component’s actual and expected losses. Separate ratios explain why the weights may differ.

Check Your Understanding

For each question, use the square-root plan-weight method. Amounts are hypothetical and on consistent bases; assume no discounting or mortality adjustment. Apply the cap at 1 where needed.

A plan has expected primary losses of $450,000 with a $1,800,000 primary standard, and expected excess losses of $200,000 with an $800,000 excess standard. Find both weights. (Answer: $Z_P=0.5000,\ Z_E=0.5000$)
Expected primary losses are $1,250,000 and the primary standard is $500,000. Expected excess losses are $360,000 and the excess standard is $1,440,000. Find both weights after any cap. (Answer: $Z_P=1.0000,\ Z_E=0.5000$)
Expected primary losses are $640,000 with a $1,600,000 standard; expected excess losses are $90,000 with a $360,000 standard. Find both weights to four decimal places. (Answer: $Z_P=0.6325,\ Z_E=0.5000$)
Expected primary losses are $2,400,000 with a $1,500,000 standard; actual primary losses are $2,100,000. Expected excess losses are $300,000 with a $1,200,000 standard; actual excess losses are $500,000. Find both weights and the combined credibility-weighted amount. (Answer: $Z_P=1.0000,\ Z_E=0.5000$; combined amount = $2,500,000)
A primary component has expected losses of $720,000 and a weight of 0.60. An excess component has expected losses of $200,000 and a weight of 0.25. Under the uncapped square-root relationship, find the full-credibility standard for each. (Answer: primary standard = $2,000,000; excess standard = $3,200,000)

What You'll Learn