Maintaining Stable Territory Boundaries

Why Territory Boundaries Must Be Stable

Territorial experience is comparable over time only when the geographic units being compared mean the same thing. A territory may be split, merged, or redrawn to reflect new information or administrative changes. If the recorded territory for an exposure changes, a difference in reported loss cost may reflect the new boundary rather than a change in the underlying risk.

A territory indication generally compares losses and exposures assigned to a territory. If the boundary used for one period differs from the boundary used for another, the comparison mixes two effects: a possible change in risk and a change in which locations are included. The goal is therefore to express historical experience on a consistent geographic basis before drawing conclusions about territorial differences.

Definition: A territory basis is the geographic boundary definition used to assign insured exposures and claims to territories. A common-basis comparison expresses the experience from each period using the same set of territory boundaries. A geographic crosswalk describes how locations or portions of old territories map to the territories in a revised definition.

The most precise method is to geocode each insured location and claim location, then assign each to the chosen boundary set. When location-level records are unavailable, a crosswalk can estimate how historical exposure and loss should be allocated. Such an estimate is only as reliable as its allocation assumptions and underlying geographic data.

Reconstructing Experience on a Common Basis

Choose the boundary definition that will serve as the comparison basis, and then express each period's exposure and losses on that basis. The chosen definition might be the current set of territories or another stable set suitable for the analysis. Record the boundary vintage and the geographic source so that the assignment can be reproduced.

Specify the comparison basis.
Identify the territory boundaries and their effective date. Use the same definition for every period being compared.

Map the insured locations.
Where possible, use the address or coordinates associated with each exposure and each claim. Apply the selected boundaries consistently to all periods.

Reconcile exposure and loss totals.
Confirm that the recoded amounts sum back to the original totals, allowing only for documented exclusions, rounding, or data corrections.

Compare on the recoded basis.
Calculate experience measures using the recoded losses and exposures. Interpret differences with due attention to population mix and any remaining allocation uncertainty.

When a historical territory is divided among several new territories, the best allocation is based on the actual location of each exposure and claim. If only old-territory totals are available, an exposure crosswalk may be used to estimate allocations. Applying an exposure share to losses assumes, in effect, that the loss experience is distributed across the old territory in proportion to the exposed units. That assumption can be poor when risk or claim severity varies within the old territory.

Key result: If recoding is complete and consistent, the sum of the recoded territory exposures equals the original total exposure, and the sum of the recoded territory losses equals the original total losses. A change in territory labels alone should not create or remove exposure or loss.

Why Weighted Loss Costs Reconcile

For a territory with positive exposure, define its loss cost as losses divided by exposures. If a territory is split into parts, the combined loss cost is the exposure-weighted average of the part loss costs. This follows directly from the definitions: each part's losses equal its loss cost times its exposure.

L_j = E_j C_j,\qquad C_{\mathrm{combined}} = \frac{\sum_j L_j}{\sum_j E_j} = \frac{\sum_j E_j C_j}{\sum_j E_j}

Here $E_j$ denotes exposure, $L_j$ denotes losses, and $C_j=L_j/E_j$ denotes loss cost for part $j$. The formula explains why a simple average of the part loss costs is generally wrong: it gives equal weight to parts with different exposure volumes. It also gives a useful reconciliation check after recoding.

Worked Examples

Worked Example: Recovering Experience After a Territory Split

A hypothetical insurer is revising one territory into two, P and Q. Historical experience for the original territory totals 240 exposure units and $360,000 in losses. Detailed location records identify 90 exposure units and $153,000 in losses as belonging to P. The remainder belongs to Q. Assume the losses are already on a common evaluation basis. Calculate each new territory's exposure, losses, and loss cost, and check the recoding.

Method. Subtract P's known amounts from the original totals to obtain Q's amounts, then divide each territory's losses by its exposures.

For P, the loss cost is:

C_P = \frac{\$153{,}000}{90} = \$1{,}700 \text{ per exposure unit}.

For Q, the remaining exposure is $240-90=150$ units, and the remaining losses are $\$360{,}000-\$153{,}000=\$207{,}000$. Thus:

C_Q = \frac{\$207{,}000}{150} = \$1{,}380 \text{ per exposure unit}.

Check. The exposures reconcile: $90+150=240$. The losses reconcile: $\$153{,}000+\$207{,}000=\$360{,}000$. The combined loss cost is $\$360{,}000/240=\$1{,}500$ per unit. The weighted calculation agrees: $(90\times \$1{,}700+150\times \$1{,}380)/240=\$1{,}500$ per unit.

Common mistake avoided. Using 180 as Q's exposure would give $\$207{,}000/180=\$1{,}150$, but 180 is not the remainder: $240-90=150$. The corrected Q loss cost is $1,380 per exposure unit.

Worked Example: Using an Exposure Crosswalk

A hypothetical company has historical records for old territories A and B, but not individual claim locations. Territory A has 100 exposure units and $180,000 in losses; territory B has 150 exposure units and $240,000 in losses. A geographic crosswalk estimates that 60% of A and 20% of B belong in revised territory X, with the remainder in revised territory Y. Estimate exposures, losses, and loss costs for X and Y. State the key assumption and verify the totals.

Method. Because individual claim locations are unavailable, apply each old territory's crosswalk shares to both its exposures and its losses. This is an estimate that assumes losses are distributed within each old territory in proportion to exposure.

For X, the allocated exposure is $100(0.60)+150(0.20)=60+30=90$ units. The allocated losses are $\$180{,}000(0.60)+\$240{,}000(0.20)=\$108{,}000+\$48{,}000=\$156{,}000$. Therefore:

C_X = \frac{\$156{,}000}{90} = \$1{,}733.33 \text{ per exposure unit}.

For Y, the shares are 40% of A and 80% of B. The exposure is $100(0.40)+150(0.80)=40+120=160$ units. The losses are $\$180{,}000(0.40)+\$240{,}000(0.80)=\$72{,}000+\$192{,}000=\$264{,}000$. Thus:

C_Y = \frac{\$264{,}000}{160} = \$1{,}650.00 \text{ per exposure unit}.

Check. The revised exposures total $90+160=250$, matching $100+150=250$. The revised losses total $\$156{,}000+\$264{,}000=\$420{,}000$, matching $\$180{,}000+\$240{,}000=\$420{,}000$.

Common mistake avoided. These are crosswalk-based estimates, not exact territory results. Applying exposure percentages to losses may misallocate losses if claim frequency or severity varies across the old territory. If claim addresses are available, use them to assign losses directly rather than relying on this proportional allocation.

Worked Example: Separating a Population Shift from a Risk Change

A hypothetical territory's insured population consists of two risk groups, A and B. In both of two periods, the loss costs are $1,000 per exposure for A and $3,000 per exposure for B. Period 1 has 80 A exposures and 20 B exposures. Period 2 has 20 A exposures and 80 B exposures. Calculate each period's aggregate loss cost, then compare the periods using a fixed 50% A, 50% B mix.

Method. Calculate the observed loss cost for each period, then standardize both to the same exposure mix. This separates a composition change from a change in group-specific costs.

In Period 1, losses are $80(\$1{,}000)+20(\$3{,}000)=\$80{,}000+\$60{,}000=\$140{,}000$, over 100 exposures. The aggregate loss cost is $\$140{,}000/100=\$1{,}400$ per exposure.

In Period 2, losses are $20(\$1{,}000)+80(\$3{,}000)=\$20{,}000+\$240{,}000=\$260{,}000$, also over 100 exposures. The aggregate loss cost is $\$260{,}000/100=\$2{,}600$ per exposure.

For the common 50%/50% mix, each standardized population has 50 exposures in each group. Its expected losses are $50(\$1{,}000)+50(\$3{,}000)=\$200{,}000$ over 100 exposures. The standardized loss cost is therefore $2,000 per exposure in both periods.

Check and interpretation. The observed aggregate loss cost rises by $1,200 per exposure, but the two group-specific costs have not changed. The common-mix comparison is unchanged at $2,000 per exposure. In this example, the observed change is entirely explained by the shift toward group B, which has the higher loss cost. A common mix does not prove that all underlying risks are unchanged; it only controls the mix of the groups included in the standardization.

Common mistake avoided. Treating the increase from $1,400 to $2,600 as evidence that the territory itself became riskier ignores the substantial change in population mix.

Practical Choices and Common Traps

Boundary consistency and population consistency are related but distinct. Recoding puts experience on the same geographic map. Mix standardization addresses changes in the types of risks within those mapped territories. A comparison can be geographically consistent and still be affected by a shift in policyholder, vehicle, property, or coverage mix.

When selecting a recoding approach, consider the following:

Prefer location-level assignment. Assign each exposure and claim to a boundary using its location when that information is reliable and available. Confirm that the location refers to the insured risk, not merely to a mailing or processing address.
Document crosswalk assumptions. Record the source, effective date, and allocation basis. A crosswalk based on area, population, or exposure counts may not represent the distribution of insured risk or losses.
Keep exposure and loss assignment coherent. Do not assign exposures using one geographic basis and losses using a different one without explaining why. A claim should generally be associated with the location of the insured risk that generated it.
Check conservation. Reconcile total exposures and losses before and after recoding. Investigate unexplained differences rather than treating them as harmless boundary effects.
Distinguish real change from coding change. If a policy's recorded territory changes because a boundary was revised, that alone is not evidence that its risk changed. Conversely, genuine changes in the population or risk mix should not be concealed by recoding.

Key takeaway: Compare territorial experience on a common boundary basis, reconcile the recoded totals, and make allocation uncertainty explicit. Then assess population-mix changes separately; a stable map does not guarantee a stable insured population.

Check Your Understanding

For each problem, identify the relevant geographic or mix issue and calculate the requested result.

A hypothetical old territory has 200 exposure units and $310,000 in losses. After a boundary revision, territory X is known to contain 70 units and $126,000 in losses. Find the remaining territory's exposure and loss cost, and check the total losses. (Answer: 130 exposure units; $1,415.38 per unit; total losses reconcile to $310,000.)
Old territory A has 80 units and $120,000 in losses; old territory B has 120 units and $180,000 in losses. A crosswalk assigns 75% of A and 25% of B to revised territory N, and the remainder to S. Allocate both exposures and losses in the same proportions, then calculate N and S loss costs. (Answer: N has 90 units and $135,000 in losses, or $1,500 per unit; S has 110 units and $165,000 in losses, or $1,500 per unit.)
A territory has group-specific loss costs of $900 for group A and $2,100 for group B. Period 1 has 70 A and 30 B exposures; Period 2 has 30 A and 70 B exposures. Calculate both observed aggregate loss costs and the common-mix cost using 50% of each group. (Answer: Period 1, $1,260 per unit; Period 2, $1,740 per unit; common-mix cost, $1,500 per unit in each period.)
A crosswalk allocates exposures from an old territory 40% to X and 60% to Y. The company has claim-level locations but no evidence that loss costs are uniform within the old territory. Which basis should be preferred for assigning claims, and why? (Answer: Assign claims using their claim or insured-risk locations and the revised boundaries; applying the exposure percentages to losses would impose an unsupported proportional-loss assumption.)
Two periods have been recoded to the same boundaries, and each has the same total exposure. The later period has a higher aggregate loss cost, but its mix contains more of a group with historically higher loss cost. What additional comparison helps distinguish mix effects from a change in risk, and what must be held consistent? (Answer: Standardize or reweight both periods to the same group mix, holding the group definitions and weights consistent.)

What You'll Learn