Calculating Square-Root Plan Weights

Purpose of the Square-Root Weight

A credibility weight expresses how much weight a rating plan assigns to an experience component relative to its complement. In this tutorial, the weight is determined by the expected loss volume associated with that component and a plan-specific full-credibility threshold. A larger expected loss volume generally means a more stable experience signal, but the weight does not increase in direct proportion to expected losses.

The calculation depends on using the expected losses for the correct experience period, coverage, and loss component. Expected losses are not the same as the actual losses observed in that period. The expected amount is calculated on the plan's specified basis, using the applicable exposures and expected loss rates; the threshold must be on that same basis.

Definition: Let $E$ be the expected losses associated with the experience component being weighted, and let $E_{\text{full}}$ be the positive expected-loss volume at which that plan assigns full credibility to the component. The plan-specific square-root weight is $$ Z=\min\left(1,\sqrt{\frac{E}{E_{\text{full}}}}\right). $$ Here $Z$ is between 0 and 1. The threshold $E_{\text{full}}$ is a plan input; this calculation uses the threshold rather than deriving it.

Both $E$ and $E_{\text{full}}$ must use matching units and definitions. For example, if the threshold applies to expected losses for a particular coverage over a defined experience period, do not compare it with expected losses for a different period or with an unrelated total. A plan may specify other adjustments to the expected-loss basis; apply those before calculating $Z$.

Calculating the Weight

First determine expected losses for the relevant experience component. If exposures are stated in units with an expected loss rate per unit, expected losses are the sum of each exposure amount multiplied by its matching rate. If the component combines classes or periods, add the expected losses only after putting them on the plan's consistent basis. Then divide by the full-credibility threshold, take the square root, and apply the cap at 1.

Identify the component and basis.
Specify which expected losses are being weighted, including the coverage and experience period used by the plan.

Calculate $E$.
Use the applicable exposures and expected loss rates, or sum the plan-consistent expected loss amounts supplied in the problem.

Use the matching threshold.
Identify the plan's $E_{\text{full}}$ for this component and basis.

Calculate and cap $Z$.
Evaluate $\sqrt{E/E_{\text{full}}}$. If the result exceeds 1, set $Z=1$; do not assign more than full credibility.

Worked Example: Weight From Exposure and a Loss Rate

A hypothetical property plan has 240 insured units in the experience period. The expected loss rate is $185 per unit on the applicable coverage and period basis. The plan's full-credibility threshold for this component is $100,000 of expected losses. Calculate the square-root weight.

Method. Calculate the expected loss volume from exposure and the expected loss rate, then apply the plan-specific square-root rule. The amounts are in dollars, and no intermediate rounding is needed.

Step 1: Calculate expected losses. There are 240 units at $185 per unit:

$$ E=240(185)=44{,}400. $$

Thus $E=\$44{,}400$, while $E_{\text{full}}=\$100{,}000$. Both amounts refer to the same hypothetical coverage and experience period.

Step 2: Divide by the threshold.

\frac{E}{E_{\text{full}}} =\frac{44{,}400}{100{,}000} =0.444.

Step 3: Take the square root and cap. Since $\sqrt{0.444}=0.666333\ldots$, this value is below 1, so the cap does not change it. Rounded to four decimal places, $Z=0.6663$.

Check. Expected losses are less than the full-credibility threshold, so the weight should be below 1. The calculated weight, 0.6663, satisfies that condition. A common mistake is to use 240 as $E$; 240 is the exposure count, not the dollar amount of expected losses.

Why the Square Root Appears

The square-root relationship can be justified under a common model for aggregate losses. Suppose the aggregate loss amount $L$ for the component has expected value $E[L]=E$ and variance $\operatorname{Var}(L)=kE$, where $k>0$ is constant across the experience volumes being compared. This assumption says variance grows in proportion to expected losses. It is an assumption used to motivate the rule, not a guarantee that every loss process has this variance structure.

For a positive expected loss amount $E>0$, the coefficient of variation is the standard deviation divided by the mean. The assumption gives:

\operatorname{CV}(L) =\frac{\sqrt{\operatorname{Var}(L)}}{E[L]} =\frac{\sqrt{kE}}{E} =\sqrt{\frac{k}{E}}.

This derivation requires $E>0$, so that the mean in the denominator is nonzero. Comparing a positive-volume component with the full-credibility volume $E_{\text{full}}>0$, the relative stability is represented by the ratio of their coefficients of variation:

\frac{\operatorname{CV}(L\text{ at }E_{\text{full}})} {\operatorname{CV}(L\text{ at }E)} = \frac{\sqrt{k/E_{\text{full}}}}{\sqrt{k/E}} = \sqrt{\frac{E}{E_{\text{full}}}}.

The positive constant $k$ cancels. This ratio is the uncapped square-root weight when $0<E<E_{\text{full}}$. At $E\geq E_{\text{full}}$, the raw ratio is at least 1, but the plan assigns no more than full credibility, which explains the minimum with 1 in the formula.

If $E=0$, the coefficient of variation is undefined because $E[L]=0$. The derivation above therefore does not apply at zero. The plan definition handles this case directly: for $E_{\text{full}}>0$, the square-root expression is $\sqrt{0/E_{\text{full}}}=0$, so $Z=0$. It is also the continuous extension of the positive-$E$ weight as $E$ approaches zero. If the variance is not proportional to expected losses, or if the plan specifies a different weight rule, this square-root justification need not apply.

Applying the Full-Credibility Cap

The cap is essential: the square-root expression describes increasing weight below the threshold, but once the experience component meets or exceeds the plan's full-credibility volume, the weight remains 1. Expected losses above the threshold do not create a weight greater than full credibility under this rule.

Worked Example: Combining Classes and Applying the Cap

A hypothetical liability plan combines two classes for one experience component. Class A has 320 exposures and an expected loss rate of $150 per exposure. Class B has 180 exposures and an expected loss rate of $200 per exposure. The plan's full-credibility threshold for the combined component is $75,000. Calculate $E$ and $Z$.

Method. Calculate each class's expected losses, add them on the stated common basis, and compare the combined amount with the plan threshold.

Step 1: Calculate Class A expected losses.

E_A=320(150)=48{,}000.

Step 2: Calculate Class B expected losses.

E_B=180(200)=36{,}000.

Step 3: Aggregate the expected losses.

E=E_A+E_B=48{,}000+36{,}000=84{,}000.

Step 4: Calculate the raw square-root value. With $E_{\text{full}}=\$75{,}000$:

\sqrt{\frac{E}{E_{\text{full}}}} =\sqrt{\frac{84{,}000}{75{,}000}} =\sqrt{1.12} =1.0583\ldots.

This exceeds 1, so the plan cap applies and $Z=1$. The final weight is 1.0000 when shown to four decimal places.

Check. The combined expected losses of $84,000 exceed the $75,000 threshold, so the final weight must be full credibility and cannot exceed 1. A common mistake is to report 1.0583 as the credibility weight rather than as the uncapped intermediate result.

Zero Volume and the Correct Loss Basis

A zero expected loss amount should not be treated as a small positive amount when calculating the weight. Provided $E_{\text{full}}>0$, the plan rule gives $Z=0$. This does not mean that the underlying risk has no possible losses; it means the expected loss volume assigned to this particular experience component is zero under the stated inputs.

Worked Example: Zero Expected Loss Volume

A hypothetical specialty coverage has no exposures in the specified experience period. The expected loss rate is $240 per exposure, and the plan's positive full-credibility threshold is $36,000. Calculate the weight for this period.

Method. Calculate expected losses first. Do not try to calculate a coefficient of variation at zero volume.

Step 1: Calculate expected losses. With zero exposures:

$$ E=0(240)=0. $$

Step 2: Apply the plan rule. Here $E_{\text{full}}=\$36{,}000>0$, so:

Z=\min\left(1,\sqrt{\frac{0}{36{,}000}}\right) =\min(1,0) =0.

The weight is 0.0000. The coefficient of variation is not calculated: with zero expected losses, its denominator would be zero. The direct plan formula remains well-defined and gives the answer.

Check. A component with no expected loss volume receives zero weight under the stated rule. The common mistake is to use the coefficient-of-variation derivation at $E=0$; that derivation requires $E>0$.

Common Exam Traps

The calculation is short, but the setup can be easy to misread. In particular, the numerator is expected losses, not actual losses, and the threshold is specific to the plan component. Keep these checks in view:

Do not reverse the ratio. The rule uses $E/E_{\text{full}}$, not $E_{\text{full}}/E$. Below the threshold, reversing the ratio would incorrectly produce a value above 1.
Do not omit the cap. If $E>E_{\text{full}}$, report $Z=1$, not a weight above 1.
Do not substitute actual losses for $E$. Actual losses may be relevant elsewhere in an experience rating plan, but this weight calculation uses expected losses.
Do not divide by zero in a coefficient-of-variation argument. Handle $E=0$ directly using the plan formula.
Do not mix bases. Compare expected losses and the threshold for the same coverage, period, and component, in consistent units.

For instance, if $E=\$36{,}000$ and $E_{\text{full}}=\$64{,}000$, the incorrect reversed calculation is $\sqrt{64{,}000/36{,}000}=1.3333\ldots$, which might then be capped at 1. The correct calculation is:

Z=\sqrt{\frac{36{,}000}{64{,}000}} =\sqrt{0.5625} =0.75.

The correct weight is 0.7500, not 1.0000. The cap is not a substitute for using the correct ratio.

Check Your Understanding

For each problem, use the expected-loss basis and full-credibility threshold stated in the question.

A hypothetical commercial package has 50 exposures at an expected loss rate of $420 per exposure. Its full-credibility threshold is $84,000. Calculate $E$ and $Z$. (Answer: $E=\$21{,}000$, $Z=0.5000$)
A hypothetical equipment plan combines 120 exposures at $250 expected loss per exposure and 80 exposures at $375 per exposure. The threshold for the combined component is $90,000. Calculate the combined weight. (Answer: $E=\$60{,}000$, $Z=0.8165$)
A hypothetical coverage has expected losses of $130,000 on the plan's specified basis and a full-credibility threshold of $100,000. What is the final weight after applying the cap? (Answer: $Z=1.0000$)
A hypothetical experience component has zero expected loss volume and a positive full-credibility threshold of $25,000. State the weight, and indicate whether a coefficient of variation is needed. (Answer: $Z=0.0000$; no coefficient of variation is calculated)
A hypothetical fleet has 400 exposures at $90 expected loss per exposure and 150 exposures at $160 per exposure. The plan threshold is $150,000. Calculate total expected losses and the square-root weight, rounding the weight to four decimal places. (Answer: $E=\$60{,}000$, $Z=0.6325$)

What You'll Learn