Credibility Is Not Rate Adequacy

Two Different Questions

Credibility and rate adequacy are related, but they answer different questions. Credibility asks how much weight an estimate should place on a particular body of experience. Rate adequacy asks whether the rate currently being charged is sufficient for the projected costs and other provisions on the rating basis being considered.

A highly credible estimate can be above or below the current rate basis. Credibility does not tell us whether the current rate is too high, too low, or just right; it tells us how much the experience contributes to an estimate. After forming that estimate, we must compare it with the appropriate current basis to assess the indicated change.

Definition: A credibility-weighted estimate combines an estimate from a subject group’s experience with a relevant complement. If $R_{\text{exp}}$ is the experience estimate, $R_{\text{comp}}$ is the complement, and $Z$ is the credibility assigned to the experience, then $$ R_{\text{cred}}=Z R_{\text{exp}}+(1-Z)R_{\text{comp}}, \qquad 0\leq Z\leq 1. $$ Rate adequacy is assessed by comparing the projected value on the indicated basis with the value underlying the current rate, using consistent measures and assumptions.

The values being compared must be comparable. For example, if the estimate is a projected loss cost per exposure, compare it with the current rate’s loss provision per the same exposure basis—not with the full premium rate unless that is what the calculation calls for. The examples below use a simplified setting: the current loss provision is known, and all other rating assumptions are held constant. Under that assumption, a proportional change in projected loss cost gives the same proportional change in the rate.

Why Credibility Weighting and Adequacy Are Separate

In a simple model, let $\theta$ be the unknown quantity to estimate, such as a group’s expected loss cost. Suppose the experience estimate $Y$ is unbiased for $\theta$, with process variance $\sigma^2$. Suppose the complement $m$ is the prior mean for $\theta$, and the variation of $\theta$ around that mean is $\tau^2$. Consider a linear estimate $m+Z(Y-m)$. Its mean squared error is the sum of the variation from uncertainty about $\theta$ and the process variation in $Y$: $$ \operatorname{MSE}(Z)=(1-Z)^2\tau^2+Z^2\sigma^2. $$

Differentiating with respect to $Z$ and setting the derivative equal to zero gives $$ -2(1-Z)\tau^2+2Z\sigma^2=0, $$ so $$ Z=\frac{\tau^2}{\tau^2+\sigma^2}. $$

This model illustrates the role of credibility: a more precise experience estimate (smaller $\sigma^2$) receives more weight, all else equal. The calculation uses assumptions about the experience and complement; it does not compare the resulting estimate with the current rate. That second comparison is an adequacy question.

Specify the measure.
State exactly what the experience estimate represents, such as projected loss cost per exposure, and identify the corresponding current-basis value.

Form the credibility-weighted estimate.
Apply the selected credibility $Z$ to the experience estimate and $1-Z$ to the relevant complement.

Compare with the current basis.
For a proportional loss-cost indication, divide the credibility-weighted projected loss cost by the current loss provision. Subtract 1 to express the result as a percentage change.

Key result: When the current rate’s loss provision is $L_{\text{current}}$, and the projected loss cost is $L_{\text{cred}}$, the indicated proportional change in the loss provision is $$ \frac{L_{\text{cred}}}{L_{\text{current}}}-1. $$ This ratio is an appropriate rate change only under the stated simplifying assumption that the remaining rating provisions and assumptions do not alter the proportional relationship.

Worked Comparisons

Worked Example: Highly Credible Experience Indicating an Increase

A hypothetical commercial-equipment group has a projected experience loss cost of \$620 per exposure. A relevant complement is \$580 per exposure, and the selected credibility is $Z=0.90$. The current rate basis includes a loss provision of \$560 per exposure. Assuming all other rating provisions remain unchanged and scale proportionally, calculate the indicated change.

Step 1: Calculate the experience weight and complement weight. The experience receives weight $0.90$, and the complement receives weight $1-0.90=0.10$.

Step 2: Calculate the credibility-weighted loss cost. $$ L_{\text{cred}}=0.90(620)+0.10(580)=558+58=\$616. $$

Step 3: Compare with the current loss provision. $$ \frac{616}{560}=1.10, \qquad 1.10-1=0.10=10.0\%. $$ The indicated change is a 10.0% increase. The experience is highly credible, and its weighted estimate is still materially above the current loss provision.

Check: A 10.0% increase to \$560 gives $\$560(1.10)=\$616$, the credibility-weighted estimate. The common mistake to avoid is treating $Z=0.90$ as evidence that the current rate is 90% adequate; it is a weight, not an adequacy percentage.

Worked Example: Highly Credible Experience Indicating a Decrease

A hypothetical inland-watercraft class has an experience estimate of \$420 per exposure. Its selected complement is \$450 per exposure, and the credibility assigned to the experience is $Z=0.95$. The current loss provision is \$470 per exposure. Assume other rating provisions and assumptions are unchanged. Calculate the indicated proportional change.

Step 1: Calculate the weights. The experience weight is $0.95$, and the complement weight is $1-0.95=0.05$.

Step 2: Blend the estimates. $$ L_{\text{cred}}=0.95(420)+0.05(450)=399+22.50=\$421.50. $$

Step 3: Calculate the indication. $$ \frac{421.50}{470}=0.8968085\ldots, \qquad 0.8968085\ldots-1=-0.1031915\ldots. $$ The indicated proportional change is a decrease of approximately 10.3%.

Check: The weighted estimate is below the current \$470 loss provision, so the direction must be a decrease. The result does not mean the experience is unreliable: $Z=0.95$ means that the experience receives substantial weight in the estimate.

Worked Example: Credibility Changes the Indication, Not the Current Rate

A hypothetical specialty class has experience of \$590 per exposure. The selected complement is \$480 per exposure, the credibility is $Z=0.75$, and the current loss provision is \$500 per exposure. Assume the proportional comparison is appropriate. Calculate the credibility-weighted indication, and compare it with the result from using experience alone.

Step 1: Calculate the weighted estimate. The complement weight is $1-0.75=0.25$, so $$ L_{\text{cred}}=0.75(590)+0.25(480)=442.50+120=\$562.50. $$

Step 2: Calculate the indication using the weighted estimate. $$ \frac{562.50}{500}-1=1.125-1=0.125=12.5\%. $$ The credibility-weighted estimate indicates a 12.5% increase.

Step 3: Compare with the unweighted experience result. Using the experience alone would give $$ \frac{590}{500}-1=0.18=18.0\%. $$ That is not the selected credibility-weighted indication. The complement moderates the estimate, reducing the indicated increase from 18.0% to 12.5%.

Check: Applying a 12.5% increase to \$500 gives $\$500(1.125)=\$562.50$. The common mistake is to call 18.0% the credibility-based indication simply because it comes from the observed experience; the stated method first blends experience and complement.

Common Traps and Practical Interpretation

The examples illustrate two independent steps: estimate the relevant projected value, then assess it against the current basis. A high credibility level does not guarantee adequacy, and a low credibility level does not establish inadequacy. Neither direction of rate change follows from $Z$ alone.

Confusing credibility with adequacy. $Z$ is the weight assigned to experience under a selected method. It is not the percentage by which a rate should change.
Comparing unlike quantities. A loss cost per exposure and a total premium rate are not interchangeable. Match the exposure basis and the measure before taking a ratio.
Using raw experience instead of the selected estimate. If the method specifies a complement, calculate the weighted estimate before deriving an indication.
Assuming that a high $Z$ means no change. The credibility-weighted estimate can differ substantially from the current rate basis even when the experience receives nearly all the weight.
Overstating what the result proves. An indicated change is conditional on the data, projections, complement, and rating assumptions used. It is not a guarantee of future results.

In an actual rate analysis, the comparison must be made on a consistent basis and account for the provisions appropriate to the analysis. The simplified ratios here isolate the distinction between credibility and adequacy; they do not replace the full rate indication process.

Key takeaway: Credibility answers, “How much weight should this experience receive?” Adequacy answers, “How does the resulting projected value compare with the current rate basis?” Calculate the estimate first, then make the comparison.

Check Your Understanding

For each question, calculate the credibility-weighted estimate before assessing the indicated change. Unless stated otherwise, assume the proportional comparison is appropriate and all other rating assumptions are unchanged.

A hypothetical class has an experience loss cost of \$630 per exposure, a complement of \$600, credibility $Z=0.80$, and a current loss provision of \$610. What is the weighted estimate and indicated change? (Answer: \$624 per exposure; 2.3% increase.)
A hypothetical specialty group has an experience estimate of \$455, a complement of \$500, credibility $Z=0.95$, and a current loss provision of \$480. What is the weighted estimate and indicated change, rounded to one decimal percentage point? (Answer: \$457.25 per exposure; 4.7% decrease.)
A hypothetical class has experience of \$700, a complement of \$600, credibility $Z=0.60$, and a current loss provision of \$650. Calculate the weighted indication. (Answer: \$660 per exposure; 1.5% increase.)
A hypothetical group’s experience is \$530 per exposure, its complement is \$500, credibility is $Z=0.90$, and the current loss provision is \$500. A reviewer proposes a 6.0% increase based on the raw experience. What indication follows the selected credibility method instead? (Answer: \$527 per exposure; 5.4% increase.)
A hypothetical class has full credibility, an experience estimate of \$480 per exposure, and a current loss provision of \$500. Does full credibility imply that the current provision is adequate? Calculate the indication and state what the credibility result does establish. (Answer: 4.0% decrease; full credibility assigns the experience estimate full weight, but does not establish adequacy.)

What You'll Learn