Partial Credibility Below the Full Standard

Introduction

A full-credibility standard describes an amount of experience considered sufficient to meet a stated statistical criterion. An account or segment may have less experience than that standard, but its observed results can still contribute to a rating indication. Partial credibility expresses how much weight the rating method assigns to those results rather than treating them as fully reliable or ignoring them entirely.

The key interpretation is simple: the credibility factor is the weight on the experience rate. The remaining weight is placed on a manual rate, which represents the rate selected from a broader rating basis. This tutorial focuses on interpreting and using that partial-credibility weight. The method that produces a particular weight must be stated or supplied; being below a full standard alone does not determine one unique value of partial credibility.

Definition: Let $N$ denote the observed exposure measured in the units used by a stated full-credibility standard, and let $N_0$ be the exposure required for full credibility under that standard. When $N<N_0$, the experience is below the full standard. A partial credibility factor $Z$, with $0\leq Z<1$, is the weight assigned to the experience in the indicated rate. Under the usual linear credibility rating convention, if $R$ is the experience rate and $M$ is the manual rate, then $R_{\mathrm{ind}}=ZR+(1-Z)M$.

What Partial Credibility Means in a Rating

In the formula, $R$ is the rate calculated from the experience being rated, and $M$ is the manual rate used as the complementary basis. The factor $Z$ is not a claim count, an exposure, or a probability that the experience rate is correct. It is a weight in a specified rating rule. For example, $Z=0.30$ means that 30% of the indication comes from the experience rate and 70% from the manual rate.

The full-credibility standard and the partial credibility factor answer different questions. The standard identifies the amount of experience associated with meeting a specified criterion. The factor tells the user how much weight the rating procedure gives to experience below that standard. A particular limited-fluctuation procedure may specify how $Z$ is assigned, but the value should not be inferred from $N<N_0$ alone. Different procedures, standards, or rating plans may assign different weights.

The two endpoints help check the interpretation. If $Z=0$, then $R_{\mathrm{ind}}=M$, so the indication uses none of the experience rate. If $Z=1$, then $R_{\mathrm{ind}}=R$, so it uses only the experience rate. For partial credibility, $Z$ lies between these endpoints. In the examples below, the applicable rating procedure supplies $Z$; the calculations explain how to use it, rather than deriving a rule for selecting it.

R_{\mathrm{ind}}=ZR+(1-Z)M

The weights add to one: $Z+(1-Z)=1$. Consequently, the indicated rate is a weighted average of the experience rate and the manual rate. If both rates are expressed in the same units and $0\leq Z\leq1$, the indication lies between them. That is a useful check on arithmetic and on whether the weights have been applied to the intended rates.

Why the Blend Shrinks the Experience Difference

The rating formula can be rearranged to show directly what partial credibility does to the experience rate's difference from the manual rate:

\begin{aligned} R_{\mathrm{ind}} &=ZR+(1-Z)M\\ &=ZR+M-ZM\\ &=M+Z(R-M). \end{aligned}

Thus the difference between the indication and the manual rate is

R_{\mathrm{ind}}-M=Z(R-M).

This is the reason for the blend: it retains a fraction $Z$ of the experience rate's departure from the manual rate. If experience is above the manual rate, the indication is above the manual rate, but by only the fraction $Z$ of the original difference. If experience is below the manual rate, the indication is below it by the same fraction. This explanation justifies the interpretation of $Z$ without assuming that the observed experience should receive full weight.

Important condition: Use the blend only when the rating convention specifies a linear credibility indication and the experience rate and manual rate have matching definitions and units. If the quantities are not comparable, first reconcile their bases or use the rating method specified for the problem. A partial-credibility weight by itself does not define how an experience rate or a manual rate is calculated.

Worked Applications

A careful calculation separates three tasks: confirm the exposure is below the stated standard, apply the supplied credibility factor to the two rates, and then use the indicated rate to calculate any dollar total. Keeping these steps distinct prevents an exposure amount from being confused with a credibility weight.

Worked Example: Applying a Supplied Partial Weight

A hypothetical group has 36 exposure units, below its stated full-credibility standard of 144 exposure units. The applicable rating procedure assigns $Z=0.40$. The experience rate is $1,250 per exposure unit, and the manual rate is $1,100 per exposure unit. Find the indicated rate and explain the contribution of each basis.

Method and below-standard check. Compare the exposure with the standard: $36<144$, so this group is below the full-credibility standard. Use the supplied partial weight; do not try to infer another value from the exposure.

Experience contribution. The weight on experience is $Z=0.40$, so its weighted contribution is $0.40($1,250)=$500$ per exposure unit.

Manual contribution. The complementary weight is $1-Z=1-0.40=0.60$. Its contribution is $0.60($1,100)=$660$ per exposure unit.

Indication. Add the contributions: $R_{\mathrm{ind}}=$500+$660=$1,160$ per exposure unit.

Check and common mistake avoided. The answer lies between $1,100 and $1,250, as a weighted average should. Reversing the weights would give $0.60($1,250)+0.40($1,100)=$1,190$, which is not the indication for the stated $Z=0.40$. The factor $Z$ weights experience, not the manual rate.

Worked Example: Converting an Indicated Rate to a Total

A hypothetical portfolio has 48 exposure units, compared with a full-credibility standard of 120 units. A rating plan supplies $Z=0.60$. Its experience rate is $940 per unit, and its manual rate is $1,020 per unit. Calculate the indicated rate and the indicated total for all 48 units.

Method and standard check. Since $48<120$, the exposure is below the full standard. Use the given $Z$, with $1-Z=0.40$.

Blend the rates. The weighted experience contribution is $0.60($940)=$564$ per unit. The weighted manual contribution is $0.40($1,020)=$408$ per unit. Therefore, $R_{\mathrm{ind}}=$564+$408=$972$ per unit.

Calculate the indicated total. Multiply the indicated rate by the 48 exposure units: $48($972)=$46,656$.

Check and common mistake avoided. The indicated rate, $972, is between $940 and $1,020. The total is $46,656. Multiplying the experience rate by exposure and then applying $Z$ to that total, while leaving the manual total unweighted, would omit the complementary manual contribution. The same blend can be applied to totals, but both totals must receive their corresponding weights: $0.60(48($940))+0.40(48($1,020))=$46,656$.

Worked Example: Combining Two Below-Standard Segments

Two hypothetical segments are each below their respective full-credibility standards. Segment A has 30 exposure units, an experience rate of $900 per unit, a manual rate of $800 per unit, and $Z=0.20$. Segment B has 70 exposure units, an experience rate of $600 per unit, a manual rate of $700 per unit, and $Z=0.50$. Find each indicated total, the combined total, and the combined rate per unit.

Method and Segment A calculation. Apply the supplied segment-specific weight to the experience rate and its complement to the manual rate. For A, $1-Z=0.80$. The indicated rate is $0.20($900)+0.80($800)=$180+$640=$820$ per unit. The indicated total for A is $30($820)=$24,600$.

Segment B calculation. For B, $1-Z=0.50$. Its indicated rate is $0.50($600)+0.50($700)=$300+$350=$650$ per unit. Its indicated total is $70($650)=$45,500$.

Combine totals and calculate the average rate. Add the segment totals: $$24,600+$45,500=$70,100$. The combined exposure is $30+70=100$ units, so the combined rate per unit is $$70,100/100=$701$.

Check and common mistake avoided. The combined rate is the exposure-weighted rate: $[30($820)+70($650)]/100=$701$. It is not the unweighted average $($820+$650)/2=$735$, because the segments have different exposures. Also, the two segment weights are applied separately; a single average credibility factor would not in general preserve these segment indications.

Reading a Partial-Credibility Indication

An indicated rate is not a statement that the observed experience is irrelevant. Rather, it moderates how strongly that experience changes the manual basis. The algebra makes the direction and size of the change transparent. For instance, when $R>M$, the difference $R-M$ is positive, and $R_{\mathrm{ind}}-M=Z(R-M)$ is also positive for $Z>0$. The indication moves upward from the manual rate, but by less than the full experience difference when $Z<1$. When $R<M$, the same reasoning shows that the indication moves downward from the manual rate.

This interpretation matters when communicating a rating result. A rate alone does not show the weight assigned to experience; the analyst should identify the experience rate, manual rate, and credibility factor used. When exposures are divided into segments, calculate the indication for each segment on its own specified basis, convert each rate to a total when needed, and then combine totals. This preserves the meaning of the segment weights and accounts for differing exposures.

Common Exam Traps

Reversing the weights. In $ZR+(1-Z)M$, $Z$ multiplies the experience rate. The complement $1-Z$ multiplies the manual rate.
Treating below-standard exposure as a complete formula. The statement $N<N_0$ identifies experience below the standard; it does not, by itself, specify the partial factor. Use the rule or value provided.
Assuming $Z$ is the proportion of actual losses to expected losses. Here $Z$ is a credibility weight in the stated rating formula, not a loss ratio or a probability statement.
Combining segment rates without accounting for exposure. When exposures differ, an overall rate is obtained from total indicated dollars divided by total exposure, not by taking a simple average of segment rates.
Mixing units or bases. The experience and manual rates must refer to comparable exposure units and definitions before they are blended.

Key takeaway: Below the full-credibility standard, partial credibility assigns a weight $Z$ to the experience rate and $1-Z$ to the manual rate. Apply the supplied weight to the experience, verify that the resulting indication is sensible, and account for exposure when turning rates into totals or combining segments.

Check Your Understanding

For each problem, use the stated credibility factor and the linear rating convention. Treat all rates as being expressed per the exposure unit specified.

A group has 20 exposure units, below its standard of 80. The supplied factor is $Z=0.50$, the experience rate is $1,500 per unit, and the manual rate is $1,300 per unit. Find the indicated rate. (Answer: $1,400 per unit)
A below-standard account has $Z=0.25$, an experience rate of $760 per unit, and a manual rate of $840 per unit. Find the indicated rate and state whether it is above or below the manual rate. (Answer: $820 per unit, below the manual rate)
A segment has 25 exposure units, experience rate $900, manual rate $1,000, and supplied $Z=0.40$. Find the indicated rate and indicated total. (Answer: $960 per unit; $24,000 total)
Two below-standard segments have these values: A has 40 units, experience rate $1,100, manual rate $1,000, and $Z=0.25$; B has 60 units, experience rate $800, manual rate $900, and $Z=0.50$. Find the combined indicated total and rate per unit. (Answer: $92,000 total; $920 per unit)
Segment A has 10 units, experience rate $1,200, manual rate $1,000, and $Z=0.20$. Segment B has 30 units, experience rate $800, manual rate $900, and $Z=0.40$. Find the exposure-weighted combined indicated rate. (Answer: $905 per unit)

What You'll Learn