Why Adjust Territory Results for Class Mix?
A territory’s observed average loss cost depends on both its results within each class and the proportions of its exposures in those classes. Two territories can have identical class-specific loss costs but different overall averages simply because their exposure distributions differ. Conversely, a raw comparison can obscure differences in class-specific experience when the territories have different mixes.
A mix-adjusted comparison answers a controlled question: what would each territory’s average loss cost be if it had the same specified distribution of class exposures? The resulting factors compare those hypothetical averages, not the territories’ actual unadjusted averages. This is useful when the purpose is to compare territory experience on a like-for-like class basis.
The class definitions and exposure units must be comparable across territories. The loss costs should also be on the same basis—for example, consistently adjusted for development and trend if those adjustments are appropriate to the analysis. The common weights might be a selected portfolio mix or another distribution chosen for the comparison; they should not vary by territory.
Raw and Standardized Loss Costs
The raw loss cost uses each territory’s own observed class distribution. If \(E_{t,c}\) is exposure in class \(c\), then the territory’s share of exposure in that class is \(E_{t,c}/\sum_c E_{t,c}\). Thus, the raw loss cost is a weighted average of the class loss costs, with weights determined separately for each territory.
For a mix-adjusted comparison, replace each territory’s own exposure shares with the same selected weights \(w_c\). If territory \(r\) is the reference, its indicated factor is 1.000 by definition, and another territory’s factor is the ratio of the two standardized loss costs.
The factor is relative, not an absolute measure of rate adequacy. For example, a factor of 1.10 indicates a standardized loss cost 10% higher than the reference territory’s under the selected mix. It does not, by itself, establish that the territory’s current rates are too low by 10%; that conclusion would require comparing the indicated result with the current rating structure and considering other ratemaking provisions.
Why Standardization Works
The justification follows directly from weighted averaging. Suppose every territory is evaluated as though it had the same exposure proportions \(w_c\). Under that hypothetical distribution, class \(c\) contributes \(w_c\) times its class loss cost to the average, regardless of the territory’s actual exposure count in that class. Summing those contributions gives the average loss cost under the shared mix.
Choose \(w_c\) for every class, with nonnegative weights that sum to 1. These are the same weights for all territories.
Multiply each territory’s class loss cost by its corresponding common weight and add the products. This gives the loss cost that territory would have under the selected mix.
Divide the standardized loss cost for the territory being evaluated by the standardized loss cost for the reference territory. The shared weights make the comparison use the same class distribution on both sides.
The raw comparison uses different weights whenever territories have different class distributions. Standardization changes those weights to a common set but leaves each territory’s class-specific loss costs unchanged. Consequently, a change from the raw ratio to the standardized factor reflects the removal of the class-mix difference from the comparison; it is not a change in the underlying class experience.
Worked Comparisons
Worked Example: Three-Class Standardization
Two hypothetical territories have the following class loss costs and exposures. Calculate each raw loss cost, then calculate the indicated factor for Territory D relative to Territory C using a common mix of 50%, 30%, and 20% for Classes 1, 2, and 3.
| Territory | Class 1 loss cost | Class 1 exposures | Class 2 loss cost | Class 2 exposures | Class 3 loss cost | Class 3 exposures |
|---|---|---|---|---|---|---|
| C | $1,000 | 60 | $1,800 | 35 | $2,400 | 5 |
| D | $1,100 | 20 | $1,500 | 60 | $3,000 | 20 |
Method. Calculate exposure-weighted raw loss costs, then apply the same selected weights to both territories’ class loss costs. The mistake to avoid is treating the raw ratio as the mix-adjusted factor.
Territory C raw loss cost. Its total exposure is \(60+35+5=100\). Its class losses, calculated as exposure times loss cost, are \(60($1{,}000)=$60{,}000\), \(35($1{,}800)=$63{,}000\), and \(5($2{,}400)=$12{,}000\). Total losses are \($60{,}000+$63{,}000+$12{,}000=$135{,}000\), so C’s raw loss cost is \($135{,}000/100=$1{,}350\).
Territory D raw loss cost. Its total exposure is \(20+60+20=100\). Its class losses are \(20($1{,}100)=$22{,}000\), \(60($1{,}500)=$90{,}000\), and \(20($3{,}000)=$60{,}000\). Total losses are \($22{,}000+$90{,}000+$60{,}000=$172{,}000\), so D’s raw loss cost is \($172{,}000/100=$1{,}720\). The raw D-to-C ratio is \($1{,}720/$1{,}350=1.2741\), rounded to four decimals.
Standardize to the common mix. For C, the standardized loss cost is \(0.50($1{,}000)+0.30($1{,}800)+0.20($2{,}400)=$500+$540+$480=$1{,}520\). For D, it is \(0.50($1{,}100)+0.30($1{,}500)+0.20($3{,}000)=$550+$450+$600=$1{,}600\).
Indicated factor. With C as the reference, D’s factor is \($1{,}600/$1{,}520=1.0526\), rounded to four decimals. Under the selected common mix, D’s loss cost is about 5.26% higher than C’s, compared with a raw ratio about 27.41% higher.
This direction is consistent with the data: D has less exposure in Class 1 and more in both Classes 2 and 3 than C. In particular, D’s larger Class 3 share is paired with a Class 3 loss cost of $3,000, compared with C’s $2,400. The raw comparison therefore reflects both class-specific results and the different distributions; standardization isolates the comparison under the stated mix.
Check. The common weights sum to \(0.50+0.30+0.20=1\). The standardized factor is above 1 because D’s standardized loss cost, $1,600, exceeds C’s, $1,520.
Worked Example: How the Selected Mix Changes the Raw Comparison
Two hypothetical territories have two classes. Territory J has class loss costs of $900 and $1,500, with exposures of 80 and 20. Territory K has class loss costs of $1,100 and $1,700, with exposures of 30 and 70. Calculate the raw K-to-J ratio and then the indicated factor under a common mix of 40% Class 1 and 60% Class 2.
Method. First retain each territory’s actual exposure weights to calculate its raw result. Then discard those territory-specific weights for the adjusted comparison and apply the 40%/60% mix to both. The common mistake to avoid is carrying the actual exposure shares into the standardization step.
Raw results. J has 100 total exposures and losses of \(80($900)=$72{,}000\) in Class 1 and \(20($1{,}500)=$30{,}000\) in Class 2. Its raw loss cost is \(($72{,}000+$30{,}000)/100=$1{,}020\). K also has 100 exposures; its class losses are \(30($1{,}100)=$33{,}000\) and \(70($1{,}700)=$119{,}000\). Its raw loss cost is \(($33{,}000+$119{,}000)/100=$1{,}520\). Thus the raw K-to-J ratio is \($1{,}520/$1{,}020=1.4902\), rounded to four decimals.
Standardized results. J’s loss cost under the common mix is \(0.40($900)+0.60($1{,}500)=$360+$900=$1{,}260\). K’s is \(0.40($1{,}100)+0.60($1{,}700)=$440+$1{,}020=$1{,}460\). The indicated K-to-J factor is \($1{,}460/$1{,}260=1.1587\), rounded to four decimals.
Check and interpretation. The standardized factor is still above 1, consistent with K’s higher class loss cost in both classes. It is less than the raw ratio because K has a much larger share of its actual exposures in Class 2, where both territories have higher loss costs. The calculation does not say that either territory’s actual experience should be replaced by the standardized result; it answers a comparison question under one shared distribution.
Worked Example: A Mix Change Can Reverse the Relative Indication
Territory P has Class 1 and Class 2 loss costs of $1,000 and $1,800. Territory Q has corresponding loss costs of $1,200 and $1,600. With P as the reference, calculate Q’s indicated factor under (a) a 75%/25% mix and (b) a 25%/75% mix.
Method. Calculate both standardized loss costs under each distribution before forming the relative factor. The mistake to avoid is assuming that the same territory must have a higher indicated factor under every possible mix.
(a) 75% Class 1 and 25% Class 2. P’s standardized cost is \(0.75($1{,}000)+0.25($1{,}800)=$750+$450=$1{,}200\). Q’s is \(0.75($1{,}200)+0.25($1{,}600)=$900+$400=$1{,}300\). Q’s factor is \($1{,}300/$1{,}200=1.0833\), rounded to four decimals.
(b) 25% Class 1 and 75% Class 2. P’s standardized cost is \(0.25($1{,}000)+0.75($1{,}800)=$250+$1{,}350=$1{,}600\). Q’s is \(0.25($1{,}200)+0.75($1{,}600)=$300+$1{,}200=$1{,}500\). Q’s factor is \($1{,}500/$1{,}600=0.9375\).
Check and interpretation. Q’s Class 1 cost is $200 above P’s, while its Class 2 cost is $200 below P’s. The 75% Class 1 mix gives greater weight to Q’s higher-cost difference, so its factor is above 1. The 75% Class 2 mix gives greater weight to Q’s lower-cost difference, so its factor is below 1. The reversal is a consequence of the chosen mix and the class-specific differences, not an arithmetic inconsistency.
Common Mistakes and Appropriate Interpretation
A frequent error is to average class loss costs without using either the territory’s actual shares or the stated common weights. For example, in the preceding J/K comparison, an unweighted average would give J \(($900+$1{,}500)/2=$1{,}200\) and K \(($1{,}100+$1{,}700)/2=$1{,}400\), producing a factor of \(1{,}400/1{,}200=1.1667\). That is not the requested common-mix result: the specified weights are 40% and 60%, not 50% and 50%. The correct standardized costs are $1,260 and $1,460, and the correct factor is 1.1587.
Other interpretation checks matter as well. Standardization cannot compensate for inconsistent class definitions, exposure bases, or loss adjustments. Nor does a selected common mix remove differences in class experience, establish that the territories are homogeneous, or automatically account for credibility. It only holds class distribution constant in the weighted average. The resulting indication should be considered alongside the underlying experience and the broader rate-selection process.
Check Your Understanding
For each problem, calculate the requested result before checking the answer.
- Territory A has class loss costs of $800 and $1,600; Territory B has $1,000 and $1,400. Find B’s factor relative to A under a 60%/40% common mix. (Answer: A standardized cost = $1,120; B standardized cost = $1,160; factor = 1.0357.)
- Territory R has class loss costs of $900 and $1,800, with exposures of 90 and 10. Territory S has loss costs of $1,100 and $1,500, with exposures of 30 and 70. Find both raw loss costs and S’s factor relative to R under a 50%/50% common mix. (Answer: R raw = $990; S raw = $1,380; R standardized = $1,350; S standardized = $1,300; factor = 0.9630.)
- Territory M has three class loss costs of $700, $1,300, and $2,100. Territory N has $840, $1,170, and $2,310. Use common weights of 50%, 30%, and 20%, with M as the reference. Find both standardized costs and N’s factor. (Answer: M = $1,160; N = $1,233; N factor = 1.0629.)
- Territory P has Class 1 and Class 2 loss costs of $1,000 and $1,800; Territory Q has $1,200 and $1,600. Find Q’s factor relative to P under a 25%/75% mix. (Answer: P standardized = $1,600; Q standardized = $1,500; factor = 0.9375.)
- Territory X has class loss costs of $700, $1,200, and $2,500. Territory Y has $800, $1,000, and $3,000. Under common weights of 40%, 35%, and 25%, calculate Y’s factor relative to X and interpret it. (Answer: X standardized = $1,325; Y standardized = $1,420; factor = 1.0717, so Y’s standardized loss cost is about 7.17% higher.)