Common Errors in Territory Rating

Why Territory Indications Go Wrong

A territory indication is only useful when its data, comparison basis, and intended application agree. A numerical result can be calculated correctly and still be misleading: territories may contain different mixes of risks, historical premiums may reflect different rate levels, selected factors may fail to balance on the intended basis, or sparse experience may be treated as fully reliable.

This tutorial focuses on diagnosing those errors. The examples use hypothetical insurance data. Unless stated otherwise, assume losses have already been adjusted to a common level for development and trend, and that exposures and premiums cover the same period and population. Those consistency conditions matter: an error in an input cannot be repaired by a later calculation.

Definition: A territory indication is an estimate of the relative rate need for a territory, based on a defined experience period, exposure base, coverage, and comparison standard. A selected territory factor is the factor ultimately applied in the rating plan; it may differ from the raw indication after credibility, balancing, operational, or judgmental adjustments. An indication is interpretable only when its basis is stated.

Four checks catch a large share of avoidable errors: compare like exposure mixes, put premiums on a common rate level, balance factors using the stated weights, and limit the influence of experience in proportion to its credibility. Each check addresses a different problem; applying one does not substitute for the others.

Error 1: Comparing Different Exposure Mixes

A territory’s overall loss cost depends both on the loss costs of its risk groups and on the proportions of those groups in the territory. If one territory has more high-loss-cost risks, its aggregate loss cost can be higher even when its loss costs within each group are lower. A raw comparison of aggregate averages confounds risk mix with territory experience.

Let $c$ index risk groups, let $L_{tc}$ be the adjusted losses for territory $t$ and group $c$, and let $E_{tc}$ be the corresponding exposures. The group loss cost is $L_{tc}/E_{tc}$. To compare territories on a common mix, choose standard weights $w_c$, with $\sum_c w_c=1$, and calculate:

LC_t^{\mathrm{std}}=\sum_c w_c\frac{L_{tc}}{E_{tc}}.

The weights represent the same reference distribution for every territory. They should be chosen to match the intended comparison, not selected separately for each territory. This standardized result addresses mix differences; it does not by itself correct for differences in data quality, premium level, or credibility.

Worked Example: Removing a Mix Effect Before Comparing Territories

Two hypothetical territories have two risk groups. Their adjusted loss costs are as follows: Territory A has loss costs of $600 for Group 1 and $1,200 for Group 2; Territory B has $500 and $1,000, respectively. Group 2 accounts for 20% of A’s exposures but 80% of B’s. Use a common standard mix of 50% in each group. Find the raw aggregate loss costs, the standardized loss costs, and B’s relativity to A on each basis.

Method and assumptions. Use loss cost per exposure as the measure, and assume the group-level losses are comparable in coverage and adjustment basis. For a territory’s raw aggregate, weight its group loss costs by that territory’s own exposure proportions. For the standardized comparison, use the stated 50%/50% weights in both territories.

Raw Territory A. Its mix is 80% Group 1 and 20% Group 2: $0.80(600)+0.20(1{,}200)=480+240=720$.

Raw Territory B. Its mix is 20% Group 1 and 80% Group 2: $0.20(500)+0.80(1{,}000)=100+800=900$. The raw relativity is $900/720=1.25$.

Standardized Territory A. $0.50(600)+0.50(1{,}200)=300+600=900$.

Standardized Territory B. $0.50(500)+0.50(1{,}000)=250+500=750$. Its standardized relativity to A is $750/900=0.8333$, to four decimal places.

Check and common mistake. B’s raw loss cost is higher, but its group loss costs are lower in both groups. The reversal arises because B has a much larger share of Group 2. Treating the raw 1.25 ratio as a pure territory effect would be a mix-comparison error. The standardized calculation uses the same weights and yields a result consistent with the within-group comparisons.

Why Standardizing the Mix Works

The calculation is a direct weighted-average comparison. For each territory, the standardized total is built from the same set of group-level loss costs and the same weights. Therefore a difference between standardized totals cannot be caused merely by different proportions of those groups in the observed territories.

In the example, if A’s group loss costs were applied to the common mix, its standardized result would be $0.5(600)+0.5(1{,}200)=900$. Applying B’s group costs to that exact same mix gives $0.5(500)+0.5(1{,}000)=750$. Comparing 750 with 900 is an apples-to-apples comparison under that chosen mix. It does not prove that the chosen mix is appropriate for every business purpose; the standard weights must represent the portfolio or comparison basis the analysis is intended to address.

Error 2: Using Premium at an Inconsistent Rate Level

When an indication is based on a loss ratio, the premium denominator must be on a consistent rate level. Premium written or earned under old rates is not directly comparable with a target loss ratio intended for current rates. If losses are appropriately adjusted but premium is not on-leveled, the resulting loss ratio can imply too much or too little rate need.

For a simple loss-ratio indication, suppose $L$ is the adjusted loss amount, $P^{\mathrm{OL}}$ is premium adjusted to the current rate level, and $LR_{\mathrm{target}}$ is the allowable loss ratio on that same basis. The indicated multiplicative change is:

\text{Indicated factor}=\frac{L/P^{\mathrm{OL}}}{LR_{\mathrm{target}}}.

This formula assumes the loss amount and premium correspond to the same experience, coverage, and exposure base, and that the target loss ratio is appropriate for the current rating basis. It is not a replacement for adjustments to losses or for expense and profit provisions; those are reflected here only through the stipulated target loss ratio.

Worked Example: Correcting a Rate-Level Mismatch

A hypothetical territory has $480,000 of adjusted losses and $800,000 of earned premium at an older rate level. The premium on-level factor to the current rate level is 1.25. The target loss ratio is 0.60. Calculate the indicated factor and compare it with the result from incorrectly using the unadjusted premium.

Method and assumptions. Apply the premium on-level factor to earned premium, not to losses. Assume losses are already on the required development and trend basis, and the stated target loss ratio is for the current rate level.

Step 1: On-level premium. $P^{\mathrm{OL}}=800{,}000(1.25)=1{,}000{,}000$.

Step 2: Current-level loss ratio. $480{,}000/1{,}000{,}000=0.48$.

Step 3: Indicated factor. $0.48/0.60=0.80$. On the stated assumptions, this indicates a 20% decrease relative to the current rate level.

Incorrect calculation. If the old-level premium is used, the loss ratio is $480{,}000/800{,}000=0.60$, and the indicated factor is $0.60/0.60=1.00$. That result hides the rate-level mismatch and indicates no change.

Check and common mistake. The on-level premium exceeds the old premium, so the current-level loss ratio must be lower than the old-level ratio. Here it falls from 0.60 to 0.48. Applying the premium factor to losses as well, or using old-level premium against a current-level target, would not follow the stated method.

Error 3: Omitting the Balancing Weights

Territory factors are often intended to redistribute rates among territories without changing the overall rate level. To achieve that objective, factors must balance on a defined basis. A plain arithmetic average is appropriate only when equal weights are the intended basis. If territory sizes differ, normalizing by an unweighted average generally fails to preserve the weighted average rate level.

Let $r_t$ be a raw indicated factor and $w_t$ the balancing weight for territory $t$, where $w_t\geq0$ and $\sum_t w_t=1$. If the objective is a weighted average factor of 1, define the weighted mean $m=\sum_t w_t r_t$, then normalize:

r_t^{\mathrm{bal}}=\frac{r_t}{m}.

The weights must match the specified balancing objective. For instance, exposure weights and premium weights need not produce the same balance if base rates vary across territories or classes. State which basis is being preserved before interpreting the normalized factors.

Worked Example: Normalizing Factors on the Stated Exposure Basis

Two hypothetical territories have raw indicated factors of 0.80 and 1.40. Their shares of the defined balancing exposure base are 75% and 25%. Normalize the factors so their weighted average is 1.

Method and assumptions. Use the stated exposure shares as weights and preserve the weighted average factor on that basis. This is a normalization example, not a claim that exposure weighting is appropriate for every rate plan.

Step 1: Calculate the weighted mean of the raw factors. $m=0.75(0.80)+0.25(1.40)=0.60+0.35=0.95$.

Step 2: Divide each raw factor by 0.95. For Territory 1, $0.80/0.95=0.842105\ldots$. For Territory 2, $1.40/0.95=1.473684\ldots$.

Step 3: Check the weighted average. $0.75(0.842105\ldots)+0.25(1.473684\ldots)=0.631579\ldots+0.368421\ldots=1.000000$. The balanced factors, rounded to four decimal places, are 0.8421 and 1.4737.

Common mistake. The unweighted mean of the raw factors is $(0.80+1.40)/2=1.10$. Dividing by 1.10 gives approximately 0.7273 and 1.2727. Their exposure-weighted average is $0.75(0.7273)+0.25(1.2727)=0.8636$, not 1. The error is using equal weights even though the specified balancing basis assigns three times as much weight to Territory 1.

Error 4: Giving Sparse Experience Too Much Credibility

An extreme territory indication may be driven by a small and volatile sample rather than a stable difference in expected costs. Applying it at full weight can produce abrupt factor changes. Credibility methods address this by blending the territory indication with a complement, such as a broader estimate or an expected value, using a weight that reflects the reliability of the territory data.

Definition: Under a credibility-weighted approach, the selected indication is $Z$ times the territory indication plus $1-Z$ times the complement, where $0\leq Z\leq1$. The credibility $Z$ and complement must be specified for the method being used. In a square-root rule with full-credibility volume $N_0$ and observed volume $n$, $Z=\min(1,\sqrt{n/N_0})$. This rule is used only when stipulated by the analysis; it is not a universal model for every territory data set.

The blend is a weighted average, not an adjustment that should be applied twice. The volume measure, full-credibility standard, and complement must fit the experience being evaluated. The earlier material on territory credibility treats selection of such standards and complements; here the diagnostic point is to avoid silently treating sparse experience as fully credible.

Worked Example: Tempering a Sparse Territory Indication

A hypothetical territory has a raw indicated factor of 1.60. The complement is 1.05. Under the stipulated square-root rule, its experience volume is $n=36$, and the full-credibility volume is $N_0=400$. Calculate the credibility-weighted indication.

Method and assumptions. Treat the stated square-root rule and complement as appropriate for this example. Do not interpret the rule as mandatory for all territory analyses.

Step 1: Calculate credibility. $Z=\min(1,\sqrt{36/400})=\sqrt{0.09}=0.30$.

Step 2: Blend the indication and complement. $0.30(1.60)+0.70(1.05)=0.48+0.735=1.215$.

Step 3: Check the result. Since $0.30$ is between 0 and 1, the weighted result must lie between 1.05 and 1.60. The value 1.215 does. Applying the raw 1.60 indication at full weight would ignore the stipulated credibility and overstate its influence.

Common mistake. Do not blend the raw indication with the complement and then apply the same credibility a second time. The stated formula already performs the blend once.

A Practical Diagnostic Sequence

These checks are easiest to apply in a fixed order. A territory factor should not be interpreted until its inputs and comparison basis have passed the relevant checks.

Confirm comparable experience.
Check that losses and exposures refer to the same territory definitions, coverage, period, and population, and that losses have the intended development and trend adjustments.

Inspect the exposure mix.
Compare the distribution of the risk groups that affect expected cost. If the distributions differ, calculate the comparison on a common stated mix rather than treating aggregate loss costs as pure territory effects.

Align premium levels.
For a loss-ratio indication, verify that premium and the target loss ratio are on the same rate-level basis. Document any premium on-level adjustment and its direction.

Apply credibility once.
Identify the territory indication, complement, volume measure, and credibility rule. Confirm that the selected result lies between the indication and complement when $0\leq Z\leq1$.

Balance on the intended basis.
State the weights and the quantity the rating plan is intended to preserve. Recalculate the weighted average after normalization to verify the result.

A useful exam check is to ask what each adjustment is correcting. Standardizing changes the mix basis; on-leveling aligns the premium denominator; credibility changes the weight given to noisy experience; normalization enforces an aggregate constraint. Combining these operations without identifying their separate purposes can conceal errors, such as balancing factors using weights from a different portfolio than the one used to calculate the indications.

Key takeaway: A territory indication is not automatically a territory factor. Compare like exposure mixes, use premium at the intended rate level, apply credibility under a stated method, and verify the weighted balance after normalization. These diagnostics are aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Check Your Understanding

For each hypothetical situation, identify the required correction and calculate the requested result. Round only where indicated.

Territory X has loss costs of $400 and $1,000 in Groups 1 and 2. Territory Y has loss costs of $500 and $1,200. Use standard weights of 60% and 40%. Calculate each standardized loss cost and Y’s relativity to X. (Answer: X = $640; Y = $780; relativity = 1.21875, or 1.219 to three decimals)
A territory has $360,000 of adjusted losses and $600,000 of earned premium at an older level. The premium on-level factor is 1.20 and the target loss ratio is 0.60. Calculate the indicated factor using on-level premium. (Answer: on-level premium = $720,000; indicated factor = 0.8333)
Two raw factors are 0.90 and 1.30, with balancing weights 40% and 60%. Normalize them to a weighted average of 1. (Answer: weighted mean = 1.14; balanced factors = 0.7895 and 1.1404)
A raw territory indication is 1.40, its complement is 1.00, and the stipulated square-root credibility rule has $n=64$ and $N_0=256$. Calculate the credibility-weighted indication. (Answer: $Z=0.50$; selected indication = 1.20)
A territory’s raw indication is 1.30, the complement is 1.05, and the stipulated square-root rule uses $n=100$ and $N_0=400$. Calculate the credibility and blended indication, then state why using 1.30 directly would be inappropriate under this rule. (Answer: $Z=0.50$; blended indication = 1.175; using 1.30 directly assigns the sparse experience more than its stipulated credibility)

What You'll Learn