Why Grouping Exposure Records Requires Care
Exposure is often summarized by characteristics used to rate a risk, such as coverage, territory, or classification. A grouped total can show how much exposure falls in a particular segment, but only if each record is assigned to the intended group and its exposure is measured on the same basis as the other records in that group.
The central decision is the grouping key: the field or combination of fields that defines a group. A summary by territory uses territory as its key. A summary by coverage and territory uses both fields; records with different coverage-territory combinations belong to different groups. Adding another field, such as classification, creates a more detailed set of groups.
The exposure basis depends on the coverage and rating context. It may be vehicle-years, payroll measured in specified units, or insured-value-years, for example. Exposure amounts with different units are not directly additive. Even when the units match, totals can answer different questions: a coverage-specific exposure total is not necessarily a count of unique vehicles, locations, or policies.
Define the Groups Before Summing
A practical way to organize a summary is to identify the record being counted, state its exposure unit, and then specify the full grouping key. For example, “collision exposure by territory and vehicle class, measured in vehicle-years” is unambiguous. “Exposure by territory” is not complete if records for multiple coverages have been combined without deciding whether that is intended.
A cross-classification keeps each combination distinct. If the key is coverage, territory, and class, then collision–North–standard and liability–North–standard are separate groups. A summary by territory alone is a different view: it combines the records assigned to that territory across the included coverages and classes. Such a total should be interpreted according to the exposure basis and purpose of the analysis.
Here \(e_r\) is the exposure amount on record \(r\), \(k(r)\) is that record’s grouping key, and \(E_g\) is the exposure total for group \(g\). If groups form a partition of the records, their totals add back to the total exposure represented by those records:
This identity follows because every record’s exposure appears in exactly one group total. Summing the group totals therefore includes each \(e_r\) once. If a record is assigned to two groups, its exposure appears twice; if it is assigned to none, its exposure does not appear at all. This is why an apparently reasonable sum can be wrong when group definitions overlap or leave out records.
Specify what exposure the summary is intended to represent and the unit in which each amount is recorded.
List every rating attribute that defines a distinct group, such as coverage, territory, and class.
Place each record in the group matching its key, then add the exposure amounts within each group.
Check whether the groups are a partition and whether adding the groups answers the intended question.
Worked Example: Vehicle-Year Exposure by Coverage and Territory
Method. Group by the full key of coverage and territory, then sum vehicle-years within each group. A hypothetical insurer has these exposure records:
| Record | Coverage | Territory | Exposure (vehicle-years) |
|---|---|---|---|
| 1 | Collision | North | 0.40 |
| 2 | Collision | North | 0.60 |
| 3 | Collision | South | 0.50 |
| 4 | Liability | North | 0.40 |
| 5 | Liability | North | 0.60 |
| 6 | Liability | South | 0.50 |
For collision in North, add records 1 and 2: \(0.40+0.60=1.00\) vehicle-year. Collision in South has record 3, so its total is \(0.50\) vehicle-year. For liability in North, add records 4 and 5: \(0.40+0.60=1.00\) vehicle-year. Liability in South has record 6, so its total is \(0.50\) vehicle-year.
The four coverage-territory totals are therefore 1.00, 0.50, 1.00, and 0.50 vehicle-years. Their sum is \(1.00+0.50+1.00+0.50=3.00\) coverage-specific vehicle-years. As a check, the six record amounts also sum to \(0.40+0.60+0.50+0.40+0.60+0.50=3.00\).
The result is 3.00 vehicle-years across coverage records, not necessarily 3.00 unique vehicle-years. If the same vehicle has both collision and liability coverage, it can contribute exposure to each coverage-specific group. Common mistake avoided: treating the sum across coverages as exposure for unique vehicles. The correct interpretation is the total exposure represented across the included coverage records.
Keep Exposure Units Consistent
Grouping does not convert one exposure basis into another. If the records are measured in payroll dollars but the requested reporting unit is payroll per $1,000, convert each amount—or the group total—using the same divisor before reporting. State the unit in the output so that a figure such as “240” cannot be mistaken for $240 rather than 240 units of $1,000 payroll.
If the underlying records are already expressed in a standardized unit, sum those standardized amounts directly. Do not divide again. For example, converting $240,000 of payroll to units of $1,000 gives \(240{,}000/1{,}000=240\) units. A second division by 1,000 would produce the wrong scale.
Worked Example: Payroll Exposure by Territory and Classification
Method. Express each record in $1,000 payroll units, then group by territory and classification. A hypothetical book has four payroll records: North–Masonry, $240,000; North–Carpentry, $360,000; South–Masonry, $180,000; and South–Carpentry, $220,000.
Convert each amount to $1,000 units:
- North–Masonry: \(240{,}000/1{,}000=240\) units.
- North–Carpentry: \(360{,}000/1{,}000=360\) units.
- South–Masonry: \(180{,}000/1{,}000=180\) units.
- South–Carpentry: \(220{,}000/1{,}000=220\) units.
Each territory-class combination has one record, so the four group totals are 240, 360, 180, and 220 units of $1,000 payroll. A territory summary gives North \(240+360=600\) units and South \(180+220=400\) units. A separate classification summary gives Masonry \(240+180=420\) units and Carpentry \(360+220=580\) units.
Check the full total in dollars: \(240{,}000+360{,}000+180{,}000+220{,}000=\$1{,}000{,}000\). In $1,000 units, the same total is \(600+400=1{,}000\). The classification view also totals \(420+580=1{,}000\). Common mistake avoided: summing the territory totals and classification totals together. Both are separate summaries of the same records; \(600+400+420+580=2{,}000\) counts the exposure twice. Use one partition at a time.
Cross-Classified Summaries and Overlapping Views
A detailed cross-classification and a marginal summary can both be useful, but they should not be confused. A cross-classified table retains each combination of attributes. A marginal total combines across one or more attributes. Each marginal view is calculated from the underlying records; it is not an additional set of exposure to be added to the detailed table.
Some records may intentionally contribute to more than one analysis—for example, the same coverage exposure can appear in a territory summary and in a class summary. That is appropriate when the summaries are alternatives. It is not appropriate to combine those overlapping views and call the result a total for unique records. The key question is whether the groups being added form a partition for the intended measure.
Worked Example: Property Exposure With Coverage and Occupancy Groups
Method. Sum the records by coverage, territory, and occupancy; retain the stated unit of $100,000 insured-value-years. In this hypothetical example, the records are already measured in that unit:
| Coverage | Territory | Occupancy | Exposure |
|---|---|---|---|
| Building | East | Office | 4.5 |
| Building | East | Office | 2.5 |
| Building | East | Warehouse | 7.0 |
| Building | West | Office | 5.0 |
| Building | West | Warehouse | 3.0 |
| Contents | East | Office | 3.0 |
| Contents | East | Warehouse | 2.0 |
| Contents | West | Office | 1.0 |
| Contents | West | Warehouse | 2.0 |
For Building–East–Office, add the two records: \(4.5+2.5=7.0\). The other Building groups are East–Warehouse \(7.0\), West–Office \(5.0\), and West–Warehouse \(3.0\). The Building total is \(7.0+7.0+5.0+3.0=22.0\) units.
For Contents, the four group totals are East–Office \(3.0\), East–Warehouse \(2.0\), West–Office \(1.0\), and West–Warehouse \(2.0\), for a Contents total of \(3.0+2.0+1.0+2.0=8.0\) units. Across the listed coverage records, the total is \(22.0+8.0=30.0\) units.
A territory summary across both coverages gives East \(7.0+7.0+3.0+2.0=19.0\) units and West \(5.0+3.0+1.0+2.0=11.0\) units. These sum to \(30.0\), as do the detailed groups. This overall figure represents the sum of the listed building and contents exposure records; it should not be described as a count of unique properties. Common mistake avoided: adding the territory totals to the coverage totals. Both summaries cover the same records, so \(19.0+11.0+22.0+8.0=60.0\) is a double count, not a new exposure total.
Common Grouping Traps
- Leaving a rating attribute out of the key. If the intended analysis distinguishes coverage, grouping only by territory merges coverage-specific exposure.
- Mixing units. Payroll dollars and units of $1,000 payroll cannot be added as though they were the same scale. Convert consistently and label the result.
- Adding overlapping summaries. Territory totals and class totals are alternate views of the same records, not separate blocks of exposure.
- Assuming a repeated exposure is always a duplicate. One insured item may legitimately have exposure under multiple coverages. Interpret a total according to what its records represent.
- Assuming every possible group must have a positive total. A group with no assigned records contributes no exposure to the sum. A reported zero and an unrepresented group are not automatically the same data statement.
Check Your Understanding
For each problem, state the relevant grouping or calculation and preserve the exposure unit.
- A hypothetical auto portfolio has 0.25 collision vehicle-years in Central, 0.75 collision vehicle-years in Central, and 0.40 collision vehicle-years in Coastal. What is the collision exposure by territory? (Answer: Central 1.00 vehicle-year; Coastal 0.40 vehicle-year.)
- Records show $125,000 of payroll in a class and $275,000 in another class. Report total payroll in units of $1,000, then state the result in dollars. (Answer: 400 units of $1,000 payroll, or $400,000.)
- A territory summary gives 320 vehicle-years in East and 180 in West. A separate class summary gives 210 vehicle-years in Class A and 290 in Class B. What is the total exposure, and should all four figures be added? (Answer: 500 vehicle-years; no, the two summaries overlap and adding all four gives a double count.)
- In a hypothetical property file, Building exposure is 6.0 units in North and 4.0 in South; Contents exposure is 3.0 units in North and 2.0 in South. Each unit is $100,000 insured-value-years. Find the total by coverage and territory, and the total across the listed coverage records. (Answer: Building 10.0 units, Contents 5.0 units, and 15.0 coverage-record units in total.)
- Four records have exposure amounts 0.30, 0.20, 0.40, and 0.10 vehicle-years. The first three are assigned to North and the last to South. One proposed grouping also assigns the 0.30 record to South. What is wrong with the proposed group totals? (Answer: The 0.30 record is counted twice; the proposed totals sum to 1.30 vehicle-years instead of the record total of 1.00.)