From Class Experience to an Indicated Loss Cost
A class’s observed loss cost is its losses divided by its exposures, measured on a consistent basis. That observed value may be informative, but it can also be volatile when the class has limited experience. Credibility weighting moderates that volatility by blending the observed class loss cost with a collective complement: a specified loss cost used to represent the broader experience or expected level.
This tutorial focuses on calculating the weighted indication once the observed experience, credibility weight, and complement are available. The complement is treated as an input; selecting or estimating it is a separate task. The calculation is aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus.
The weight \(Z\) states how much influence to give the class’s own observed loss cost. When \(Z=1\), the indication equals the observed loss cost; when \(Z=0\), it equals the complement. For an intermediate weight, the indication lies between those two values. Neither \(Z\) nor the complement changes the definition of the observed loss cost: first calculate losses per exposure, then apply the weights.
In the basic Bühlmann setting with \(n\) comparable observations and structural parameter \(K\), the credibility weight is \(Z=n/(n+K)\). Use this expression only when \(n\) and \(K\) are measured on the same observation or exposure basis assumed by that model. In some problems the weight is supplied directly. In either case, the indicated class loss cost is calculated using the same blending formula.
Why the Weighted Formula Works
The blend is a weighted average of two estimates of the same quantity: the class’s loss cost. Its weights sum to one, since \(Z+(1-Z)=1\). Thus, if the class observation and complement are both expressed in dollars per exposure unit, their weighted average is also in dollars per exposure unit.
The same calculation can be viewed in aggregate-loss terms. Multiplying the indicated cost by \(E\) gives:
The first contribution is the credibility-weighted observed aggregate loss, \(ZL\). The second is the remaining weight applied to the complement expressed as an aggregate amount over the class’s exposures, \((1-Z)EC\). This identity explains why the exposure basis matters: the complement must be multiplied by exposures in the same units used to calculate \(L/E\).
This is an indication formula, not a claim that the class’s future losses will equal the indicated cost times exposure exactly. It gives a weighted estimate under the stated inputs. It does not by itself add expenses, profit provisions, or other premium components.
Calculation Sequence
Check that losses and exposures refer to compatible coverage, time period, valuation, and exposure units. Make any required adjustments before calculating the loss cost if the problem specifies them.
Divide aggregate losses by the corresponding exposures. Do not divide by claim count unless the requested measure is a severity rather than a loss cost.
Use the given \(Z\), or calculate it from the stated credibility model with quantities on a consistent basis.
Calculate \(Z\) times observed cost plus \(1-Z\) times the complement. Confirm that the result is between the two costs when \(0\leq Z\leq1\).
Worked Calculations
Worked Example: Applying a Supplied Credibility Weight
A hypothetical class has $1,080,000 of losses over 2,400 vehicle-years. Its credibility weight is \(Z=0.35\), and the specified collective complement is $520 per vehicle-year. Calculate the indicated class loss cost and the corresponding indicated aggregate losses for the same exposure.
Method. Calculate aggregate losses per vehicle-year, then blend that observed cost with the complement. The losses and complement are both stated per vehicle-year after conversion.
Step 1: Observed loss cost.
Step 2: Complement weight. The complement receives \(1-Z=1-0.35=0.65\).
Step 3: Blend.
Step 4: Aggregate check. The indicated aggregate loss amount for 2,400 vehicle-years is \(2{,}400(\$495.50)=\$1{,}189{,}200\). From the aggregate form, the same result is \(0.35(\$1{,}080{,}000)+0.65(2{,}400)(\$520)=\$378{,}000+\$811{,}200=\$1{,}189{,}200\).
The indication, $495.50, lies between the observed $450 and complement $520, as it should. The common error this method avoids is assigning the complement weight \(Z\) and the experience weight \(1-Z\), which would reverse the intended blend.
Worked Example: Calculating the Weight Before the Indication
A hypothetical specialty class has 3,600 exposure units and $1,260,000 of losses. For this problem, the applicable Bühlmann parameter is \(K=5,400\) exposure units. The collective complement is $470 per exposure unit. Calculate the credibility weight and indicated loss cost.
Method. The problem specifies the basic Bühlmann form \(Z=n/(n+K)\). Here, \(n\) is 3,600 exposure units, and \(K\) is also in exposure units.
Step 1: Calculate credibility.
The complement weight is \(1-Z=0.60\).
Step 2: Calculate observed loss cost.
Step 3: Calculate the indicated cost.
The answer is between $350 and $470. A frequent trap is using \(Z=K/(n+K)\), which would incorrectly produce 0.60 for the experience weight. Under the stated formula, the numerator is the class’s exposure \(n\), so its weight is 0.40.
Worked Example: Pooling Experience Across Two Periods
A hypothetical class has the following experience on a consistent coverage and valuation basis: Period 1 has 800 exposure units and $420,000 of losses; Period 2 has 1,200 exposure units and $780,000 of losses. The specified Bühlmann parameter is \(K=3,000\) exposure units, and the complement is $525 per exposure unit. Calculate the pooled observed loss cost, credibility weight, and indication.
Method. Pool by adding losses and exposures separately. Then calculate the weight using total exposure and apply the credibility blend.
Step 1: Pool losses and exposures.
Step 2: Calculate pooled observed loss cost.
For comparison, the separate period costs are \(\$420{,}000/800=\$525\) and \(\$780{,}000/1{,}200=\$650\). Their simple average is $587.50, which is not the pooled loss cost because the periods have unequal exposures.
Step 3: Calculate credibility.
Step 4: Blend.
The indication lies between $525 and $600, and the pooled observation gives more weight to Period 2 because it has more exposure. The common mistake this avoids is averaging period loss costs equally rather than dividing total losses by total exposures.
Interpretation and Common Traps
The credibility-weighted class loss cost is not a weighted average of claim counts, total losses, or class relativities. It is a weighted average of two loss costs on the same basis. Establish that basis before applying the formula.
| Check | Correct approach | Common error |
|---|---|---|
| Observed experience | Use aggregate losses divided by corresponding exposures | Use losses alone as a per-exposure cost |
| Weights | Experience receives \(Z\); complement receives \(1-Z\) | Reverse the two weights |
| Multiple periods | Pool total losses and total exposures | Take an unweighted average of period costs when exposure differs |
| Units | Express the complement per the same exposure unit as the observation | Blend costs stated per different exposure bases |
A useful numerical check is that a convex blend must lie between the observed cost and the complement. If the answer falls outside that interval, verify the weights, signs, arithmetic, and units. If \(Z=0\) or \(Z=1\), the formula should reduce to the complement or observed experience, respectively.
The calculation also depends on the stated inputs being appropriate for the intended indication. In particular, a weight based on exposure units cannot be combined with a \(K\) measured in a different unit without a justified conversion. A complement from a different coverage, currency, or cost basis must not be blended as though it were comparable. This formula does not determine how to adjust experience for trend, development, or changes in exposure; apply any such adjustments when the problem specifies them.
Check Your Understanding
For each problem, calculate the requested indication using the stated data and basis.
- A hypothetical class has $672,000 of losses over 1,400 exposure units. If \(Z=0.25\) and the complement is $560 per exposure unit, what is the indicated loss cost? (Answer: $540.00 per exposure unit)
- A class has $1,500,000 of losses over 2,500 exposure units. With \(K=7,500\) exposure units and a $520 complement, calculate \(Z\) and the indicated cost using the basic Bühlmann formula. (Answer: \(Z=0.25\); $540.00 per exposure unit)
- A hypothetical class has $330,000 of losses over 600 exposure units in one period and $540,000 over 900 units in another. If \(K=1,500\) exposure units and the complement is $620, calculate the pooled observed cost, \(Z\), and the indication. (Answer: pooled cost $580.00; \(Z=0.50\); indication $600.00 per exposure unit)
- The observed class cost is $410 per exposure unit, the specified complement is $500, and \(Z=0.70\). Calculate the indicated cost and state whether it is closer to the observation or the complement. (Answer: $437.00 per exposure unit; closer to the observation)
- A class has $430,000 of losses over 1,000 exposure units. The basic Bühlmann parameter is \(K=1,500\) exposure units, and the complement is $550 per exposure unit. Calculate the indicated cost and its implied aggregate amount over the observed exposures. (Answer: \(Z=0.40\); $502.00 per exposure unit; $502,000)