Tutorials › CAS Exam 5 - Basic Ratemaking and Estimated Claim Liabilities › Partial Credibility for Loss Ratios

Home  /  Tutorials  /  CAS Exam 5 / Classical Credibility Methods

CAS Exam 5 · Classical Credibility Methods · Tutorial 549 of 989

Partial Credibility for Loss Ratios

Calculate loss-ratio credibility from premium volume, then use it to weight observed experience against a benchmark ratio.

Intermediate 10 min read

What You'll Learn

  • Calculate the observed loss ratio from losses and fixed premium
  • Apply the square-root rule using the corresponding full-credibility premium
  • Explain why partial credibility scales with the square root of premium volume
  • Blend an observed ratio with a complement-of-credibility benchmark
  • Recognize errors caused by using a linear volume ratio or inconsistent premium bases

Why Partial Credibility for a Loss Ratio Matters

An observed loss ratio can be informative without meeting the volume standard for full credibility. A portfolio with relatively little premium may have a loss ratio that fluctuates substantially from one period to another. Limited-fluctuation credibility gives a systematic way to reduce the influence of that observation: assign it a partial weight and assign the remaining weight to a benchmark ratio.

The calculation has two distinct parts. First, determine the credibility factor from the portfolio’s premium relative to the premium required for full credibility. Then, if a credibility-weighted ratio is requested, use that factor to blend the observed ratio with the benchmark. The factor is not itself a loss ratio, and it is not the observed ratio.

Definition: Let \(L\) be the observed losses, \(P\) the fixed premium corresponding to those losses, and \(P_F\) the premium volume required for full credibility under a specified fluctuation standard. The observed loss ratio is \(LR_{\mathrm{obs}}=L/P\). For premium volume at or below the full-credibility requirement, the limited-fluctuation partial-credibility factor is \(Z=\sqrt{P/P_F}\). If \(R_0\) is a selected benchmark or complement-of-credibility ratio, the credibility-weighted ratio is \(ZLR_{\mathrm{obs}}+(1-Z)R_0\).

The full-credibility premium \(P_F\) must correspond to the same loss measure, confidence criterion, error tolerance, and exposure basis as the premium \(P\) being assessed. It may be supplied directly or determined from a full-credibility standard. The formula does not make an incompatible standard appropriate merely because both quantities are stated in dollars.

From Premium Volume to a Credibility Factor

The square-root rule is used when the underlying variability is consistent with the limited-fluctuation assumptions: risks are sufficiently homogeneous, their losses are independent or dependence is negligible, and expected aggregate loss variance grows in proportion to premium. Premium must serve as a reasonable measure of volume. The rule is not a universal formula for any statistic labelled a loss ratio.

In practical terms, the partial factor compares the available premium volume with the full-credibility premium volume. If premium is one quarter of the full standard, the factor is one half, not one quarter. If premium is one ninth of the standard, the factor is one third.

$$ Z=\sqrt{\frac{P}{P_F}}, \qquad 0\leq P\leq P_F $$

A short derivation explains the square root. Suppose the variance of aggregate losses is proportional to premium, so \(\operatorname{Var}(L)\propto P\). Since the premium is fixed, the observed loss ratio is \(L/P\), and therefore

$$ \operatorname{Var}\left(\frac{L}{P}\right) =\frac{\operatorname{Var}(L)}{P^2} \propto \frac{P}{P^2} =\frac{1}{P}. $$

Taking square roots, the standard deviation of the loss ratio is proportional to \(1/\sqrt{P}\). Compare a portfolio at premium \(P\) with one at the full-credibility premium \(P_F\). Their ratio standard deviations satisfy

$$ \frac{\operatorname{SD}(L/P)}{\operatorname{SD}(L/P_F)} =\sqrt{\frac{P_F}{P}}. $$

Thus the full-standard portfolio’s fluctuation scale, relative to the less credible portfolio’s scale, is \(\sqrt{P/P_F}\). This gives the partial factor \(Z\). The derivation depends on variance proportional to premium; material heterogeneity, dependence, or a different variance-volume relationship can make the square-root rule inappropriate without adjustment.

Worked Calculations

Worked Example: Calculate the Factor and Weighted Ratio

A hypothetical specialty portfolio has observed losses of $1,200,000 and fixed earned premium of $4,000,000. The matching full-credibility premium is $16,000,000. The selected benchmark loss ratio is 0.65. Calculate the partial-credibility factor and the credibility-weighted ratio.

Step 1: Calculate the observed loss ratio. Divide losses by the fixed premium:

$$ LR_{\mathrm{obs}}=\frac{\$1{,}200{,}000}{\$4{,}000{,}000}=0.30. $$

Step 2: Calculate the premium-volume fraction.

$$ \frac{P}{P_F} =\frac{\$4{,}000{,}000}{\$16{,}000{,}000} =0.25. $$

Step 3: Apply the square-root rule.

$$ Z=\sqrt{0.25}=0.50. $$

Step 4: Blend the two ratios. The observed ratio receives weight 0.50, and the benchmark receives the remaining weight \(1-Z=0.50\):

$$ LR_{\mathrm{weighted}} =(0.50)(0.30)+(0.50)(0.65) =0.15+0.325 =0.475. $$

The partial-credibility factor is 0.50, and the weighted ratio is 0.475, or 47.5%. As a check, the result lies halfway between 0.30 and 0.65 because the two weights are equal. This calculation avoids treating the factor as the final loss ratio.

Worked Example: Keep the Premium Basis Consistent

A hypothetical program reports $2,990,000 of losses and $5,500,000 of earned premium for a period. The corresponding full-credibility premium is $8,800,000, and the benchmark ratio is 0.58. Calculate the factor and weighted ratio. All premiums are on the same basis and correspond to the reported losses.

Step 1: Calculate the observed ratio.

$$ LR_{\mathrm{obs}} =\frac{\$2{,}990{,}000}{\$5{,}500{,}000} =0.54363636\ldots $$

Step 2: Find the available fraction of full-credibility premium.

$$ \frac{P}{P_F} =\frac{\$5{,}500{,}000}{\$8{,}800{,}000} =0.625. $$

Step 3: Calculate the factor.

$$ Z=\sqrt{0.625}=0.790569415\ldots $$

The complement weight is \(1-Z=0.209430585\ldots\). Retain the unrounded factor and observed ratio during the calculation:

$$ \begin{aligned} LR_{\mathrm{weighted}} &=(0.790569415\ldots)(0.54363636\ldots) +(0.209430585\ldots)(0.58)\\ &=0.551252\ldots \end{aligned} $$

The weighted ratio, rounded to four decimal places, is 0.5513. It is closer to the observed ratio than to the benchmark because the factor is greater than one half. This example also illustrates why the two premium figures must be comparable: changing the premium basis in either the observed ratio or the credibility fraction would change the answer.

Worked Example: Diagnose a Linear-Volume Error

A hypothetical commercial portfolio has $1,750,000 of losses, $2,500,000 of fixed premium, and a full-credibility premium requirement of $10,000,000. Its benchmark ratio is 0.66. Calculate the correct factor and weighted ratio, then compare them with the result from mistakenly using the premium fraction as the factor.

Step 1: Calculate the observed ratio and premium fraction.

$$ LR_{\mathrm{obs}}=\frac{\$1{,}750{,}000}{\$2{,}500{,}000}=0.70, \qquad \frac{P}{P_F}=\frac{\$2{,}500{,}000}{\$10{,}000{,}000}=0.25. $$

Step 2: Calculate the correct partial factor. The factor is the square root of the premium fraction:

$$ Z=\sqrt{0.25}=0.50. $$

Step 3: Calculate the correct weighted ratio.

$$ LR_{\mathrm{weighted}} =(0.50)(0.70)+(0.50)(0.66) =0.35+0.33 =0.68. $$

Common mistake and correction. Using \(Z=P/P_F=0.25\) incorrectly gives

$$ (0.25)(0.70)+(0.75)(0.66)=0.175+0.495=0.670. $$

That answer gives too little weight to the observed experience. The correction is to take the square root of the volume fraction first, giving \(Z=0.50\) and a weighted ratio of 0.680. The arithmetic check is that the correct answer is midway between 0.70 and 0.66, as equal weights require.

Interpretation, Conditions, and Common Traps

The factor describes how much weight the limited-fluctuation method assigns to the observed loss ratio. It does not state that a fraction of actual losses is credible, nor does it by itself determine the final prospective ratio. A benchmark must be specified before a weighted ratio can be calculated. If the task asks only for \(Z\), stop after calculating the square root; do not introduce a benchmark that was not requested.

The premium \(P\) should be the fixed premium associated with the losses in the observed ratio. For example, a premium measure from one period should not be paired with losses from a different period unless the problem explicitly provides a consistent adjustment. Similarly, \(P_F\) must come from the applicable full-credibility requirement, not from an unrelated claim-count or severity standard.

  • Using the linear ratio. \(P/P_F\) is the volume fraction; \(Z\) is its square root.
  • Reversing the ratio. The factor uses available premium divided by full-credibility premium, not the reciprocal.
  • Confusing factor and estimate. \(Z\) is a weight between the observed and benchmark ratios, not the blended result.
  • Mixing units or bases. Numerator and denominator premiums must use the same currency units and a compatible exposure and premium basis.
  • Premature rounding. Keep additional digits for \(Z\), the observed ratio, and the complement weight until the final answer.

Within the partial range \(0\leq P\leq P_F\), the factor increases with premium but less rapidly than premium itself. For instance, quadrupling the premium fraction multiplies \(Z\) by \(\sqrt{4}=2\), provided both volumes remain within that range. When the factor is used in a blend, increasing \(Z\) moves the weighted ratio toward the observed ratio; it does not necessarily move the estimate up or down, because that depends on whether the observed ratio is above or below the benchmark.

Key takeaway: For a loss ratio based on fixed, comparable premium, calculate the observed ratio as losses divided by premium and calculate partial credibility as the square root of premium divided by the matching full-credibility premium. If a benchmark is supplied, weight it by the complement \(1-Z\).

Check Your Understanding

For each hypothetical portfolio, calculate the requested quantity using the stated premium basis. Round factors and ratios to four decimal places only where rounding is needed.

  1. A portfolio has losses of $840,000, premium of $2,100,000, and a full-credibility premium of $8,400,000. Its benchmark ratio is 0.64. Find the partial factor and weighted ratio. (Answer: \(Z=0.5000\); weighted ratio \(=0.5200\))
  2. A portfolio has losses of $3,150,000 on $4,500,000 of premium. The matching full-credibility premium is $18,000,000 and the benchmark ratio is 0.72. Find the observed ratio, factor, and weighted ratio. (Answer: observed ratio \(=0.7000\); \(Z=0.5000\); weighted ratio \(=0.7100\))
  3. A hypothetical book has losses of $4,500,000, fixed premium of $6,250,000, full-credibility premium of $25,000,000, and benchmark ratio 0.68. Calculate the factor and weighted ratio. (Answer: \(Z=0.5000\); weighted ratio \(=0.7000\))
  4. A portfolio has losses of $1,260,000 and corresponding premium of $1,800,000. Its full-credibility premium is $7,200,000 and benchmark ratio is 0.60. A candidate uses the linear premium fraction as the credibility factor. Give the correct factor and weighted ratio, and identify the incorrect factor. (Answer: correct \(Z=0.5000\); weighted ratio \(=0.6500\); incorrect linear factor \(=0.2500\))
  5. A portfolio’s full-credibility premium requirement is $20,000,000. What premium is required for a partial factor of 0.75? If losses at that premium are $7,200,000 and the benchmark ratio is 0.70, also calculate the weighted ratio. (Answer: premium \(=\$11{,}250{,}000\); observed ratio \(=0.6400\); weighted ratio \(=0.6550\))

Previous

Next

Back to Tutorials
Clinical Biostats