Why a Credibility Weight Has a Complement
A subject estimate reflects the experience or information specific to the risk, group, or account being rated. A collective estimate reflects information from a broader group. When both are used in a blended estimate, the credibility weight \(Z\) determines the subject estimate’s share. The remaining share belongs to the collective estimate: it is the complement of credibility, \(1-Z\).
The complement is not a second, separately selected credibility factor. Once \(Z\) is specified, its complement is fixed. Together, \(Z\) and \(1-Z\) allocate all the weight in the blend. The resulting estimate depends on both input estimates, so they must measure the same quantity on a comparable basis—for example, annual expected loss cost per exposure.
For example, if \(Z=0.70\), the subject estimate receives 70% of the weight and the collective estimate receives the remaining 30%. The complement does not mean that 30% of the subject estimate is subtracted. It is a positive weight applied to \(C\).
How the Two Components Form a Blend
Each estimate contributes its value multiplied by its assigned weight. The subject contribution is \(ZS\); the collective contribution is \((1-Z)C\). Adding those contributions gives \(B\). A quick check is that the weights add to one:
When \(Z=0\), the blend is entirely collective: \(B=C\). When \(Z=1\), it is entirely subject-specific: \(B=S\). Between those endpoints, \(B\) lies between \(S\) and \(C\), provided both estimates are finite and expressed in the same units. This is a useful reasonableness check; a result outside that range usually signals an arithmetic or input-basis error.
The formula can also be rearranged to show directly how the blend moves away from the collective estimate:
This form gives a short justification for the role of the complement. Start at the collective estimate \(C\), then add the fraction \(Z\) of the difference between the subject estimate and the collective estimate. If \(S\) is close to \(C\), the adjustment is small; if \(Z=0\), there is no adjustment. The identity is algebraic, not a rule for choosing \(Z\). The credibility weight must be supplied or determined by a method appropriate to the application.
A reliable calculation keeps the components visible until the final addition. Record the subject contribution and collective contribution separately. Their weights should total 1, and their weighted values should add to the blended estimate. This makes it easier to spot a reversed weight or a missing component.
Calculating and Interpreting a Complement
Confirm that the subject and collective estimates refer to the same measure, time basis, and units.
Subtract the subject weight from 1 to obtain the collective weight, \(1-Z\).
Multiply \(S\) by \(Z\) and \(C\) by \(1-Z\); do not apply either weight to the wrong estimate.
Add the contributions. Confirm the weights sum to 1 and, when \(0 \leq Z \leq 1\), that the answer lies between the two input estimates.
Worked Example: Blending Two Annual Loss Costs
A hypothetical fleet has a subject annual loss-cost estimate of $720 per vehicle. The collective estimate for comparable vehicles is $600 per vehicle. The assigned credibility weight is \(Z=0.75\). Calculate the blended annual loss cost and identify the contribution from each estimate.
Method. Use the complementary-weight blend. Both inputs are annual loss costs per vehicle, so they are on a common basis. The collective weight is $$ 1-Z=1-0.75=0.25. $$
The subject contribution is $$ ZS=0.75(720)=540. $$ The collective contribution is $$ (1-Z)C=0.25(600)=150. $$ Therefore, $$ B=540+150=690. $$
The blended annual loss cost is $690 per vehicle. The weights check because \(0.75+0.25=1.00\). The answer also lies between $600 and $720, as expected. No rounding was needed.
Common mistake avoided: Applying 0.75 to the collective estimate would give the wrong allocation. The specified \(Z\) weights the subject estimate; its complement, 0.25, weights the collective estimate.
What the Complement Does—and Does Not Do
The complement describes the collective estimate’s share of the blend. It does not describe the difference between the estimates, the percentage change in the subject estimate, or the proportion of the collective data used in some separate calculation. Its meaning is tied to the stated blend.
The two estimates also need a consistent definition. An annual loss cost should not be blended directly with an annual premium unless the premium has been put on a comparable basis. Likewise, estimates from different coverage limits or different time periods should not be combined without first addressing those differences. A mathematically correct complement cannot repair incompatible inputs.
The condition \(0 \leq Z \leq 1\) ensures both weights are nonnegative. Under that condition, the blend is a convex combination and stays between the estimates. If a calculation uses a weight outside this range, the formula may still produce a number, but the result is no longer an ordinary blend: one component has a negative weight or the other exceeds 100%. Do not describe such a result as a standard complementary-weight blend without a specific model that justifies it.
Worked Example: Converting a Blended Rate to an Amount
A hypothetical small commercial program has a subject annual loss-cost estimate of $84 per exposure and a collective estimate of $72 per exposure. The subject credibility weight is \(Z=0.40\). The program is expected to have 1,250 exposures. Find the blended annual loss cost per exposure and the corresponding total expected loss cost for those exposures.
Method. First calculate the complement: $$ 1-Z=1-0.40=0.60. $$ The subject contribution per exposure is $$ 0.40(84)=33.60. $$ The collective contribution per exposure is $$ 0.60(72)=43.20. $$ The blended annual loss cost per exposure is $$ B=33.60+43.20=76.80. $$
To obtain the total expected loss cost for 1,250 exposures, multiply the per-exposure estimate by the exposure count: $$ 1{,}250(76.80)=96{,}000. $$ Thus the blended estimate is $76.80 per exposure, corresponding to $96,000 for 1,250 exposures. The weights sum to \(0.40+0.60=1.00\), and $76.80 falls between the two input estimates. No rounding was needed.
Common mistake avoided: The exposure count is applied after blending the per-exposure estimates. Applying 1,250 to only one component, or blending a total amount with a per-exposure rate, would mix units or omit part of the estimate.
Worked Example: A Low Subject Weight
A hypothetical specialty coverage class has a subject estimate of $1,240 per policy-year and a collective estimate of $1,000 per policy-year. The assigned subject credibility weight is \(Z=0.20\). Calculate the blend and explain why it is closer to the collective estimate.
Method. The collective weight is $$ 1-Z=1-0.20=0.80. $$ The weighted subject contribution is $$ 0.20(1{,}240)=248. $$ The weighted collective contribution is $$ 0.80(1{,}000)=800. $$ Adding the components gives $$ B=248+800=1{,}048. $$
The blended estimate is $1,048 per policy-year. It is closer to $1,000 than to $1,240 because 80% of the blend’s weight is assigned to the collective estimate. As a check, the contributions sum to $1,048, the weights sum to 1, and the result lies between the two inputs. No rounding was needed.
Common mistake avoided: A small \(Z\) does not mean that the collective estimate gets a small weight. Here \(Z=0.20\), so the complement is \(0.80\), and the collective estimate receives most of the weight.
Common Weighting Errors
The complement is simple to calculate, but exam calculations can still go wrong when the roles of the two weights are not stated explicitly. Use the following checks before accepting a result:
- Do not swap the weights. In this convention, \(Z\) multiplies the subject estimate, and \(1-Z\) multiplies the collective estimate.
- Do not omit the collective contribution. The formula has two weighted terms; it is not just \(ZS\).
- Do not treat the complement as a subtraction from an estimate. Calculate \(1-Z\), then multiply that weight by \(C\).
- Check units before combining values. Both estimates must describe the same quantity on a comparable basis.
- Check the range. With \(0 \leq Z \leq 1\), the result should be between the two input estimates.
For example, with \(Z=0.40\), \(S=84\), and \(C=72\), the incorrect calculation \(0.40(84)+0.40(72)=62.40\) uses weights that sum to only 0.80. The correction is to use \(1-0.40=0.60\) on the collective estimate: \(0.40(84)+0.60(72)=33.60+43.20=76.80\). The error is not a rounding issue; it leaves part of the total weight unassigned.
Check Your Understanding
For each hypothetical situation, calculate or interpret the complementary-weight blend. Check that the stated weights are assigned to the correct estimates.
- A subject annual cost estimate is $510 per unit, the collective estimate is $450 per unit, and \(Z=0.60\). What is the complement and the blended estimate? (Answer: Complement \(0.40\); blended estimate $486 per unit.)
- A hypothetical account has a subject estimate of $2,400 per policy-year and a collective estimate of $2,000 per policy-year. If the collective weight is 0.35, find \(Z\) and the blend. (Answer: \(Z=0.65\); blended estimate $2,260 per policy-year.)
- A program’s subject estimate is $96 per exposure, its collective estimate is $80 per exposure, and \(Z=0.25\). Find the blend and the total amount for 500 exposures. (Answer: $84 per exposure; $42,000 total.)
- A blended estimate is formed from a subject value of $700 and a collective value of $640, with \(Z=0.50\). Calculate both weighted contributions and the blend. (Answer: Subject contribution $350; collective contribution $320; blend $670.)
- A subject estimate is $1,100 per contract-year and a collective estimate is $900 per contract-year. An analyst reports a blend of $1,000 using \(Z=0.70\). Is that result correct? Calculate the correct blend. (Answer: No; the complement is 0.30 and the correct blend is $1,040 per contract-year.)