Handling Midterm Changes in Exposure

Why Midterm Changes Need Careful Treatment

An experience period rarely consists of one unchanged risk from beginning to end. An endorsement may add a vehicle or change a classification; a business may begin a new operation; or the policy may be cancelled before its scheduled expiration. If the entire period is treated as though the final policy terms applied throughout, the experience record can pair losses with the wrong exposure or include time when coverage was not in force.

The basic approach is to divide the experience period wherever a material change takes effect. For each segment, record the dates, the exposure basis and amount, the insured operation or classification, and the coverage terms in force. Then attribute each loss according to the date of the event that produced it and the terms applicable on that date. This procedure supports consistent experience rating; it does not by itself determine a credibility weight or a premium adjustment.

Definition: A midterm exposure change is a change during a policy term that affects the amount, basis, or classification of exposure. Segmented earned exposure is the exposure accumulated separately over each interval in which the relevant policy terms and risk characteristics are treated as unchanged. A loss is attributed to the segment in which its covered occurrence took place, subject to the applicable coverage terms.

A segment boundary is generally the effective date of an endorsement, cancellation, change in covered operations, or other material change. The segment before the change ends immediately before that date; the new segment begins on the effective date. Policy wording controls whether coverage is in force on a boundary date, so dates should not be assigned by assumption when the documents specify otherwise.

Segment Exposure, Then Match Losses

The exposure basis should remain consistent with the coverage and measurement approach used for the experience period. It might be payroll, sales, vehicle-years, or another defined unit. When exposure is earned uniformly over time, a partial-period amount can be calculated by multiplying the full-period exposure rate by the covered fraction of the period. When actual auditable exposure is available, such as actual payroll by classification, use the applicable actual amount rather than imposing a time allocation that does not reflect the underlying exposure.

E_{\text{period}}=\sum_{j=1}^{m} E_j

Here, $E_j$ is the exposure earned in segment $j$, and the segments must cover the intended experience period without overlap or gaps in covered time. This sum is meaningful only when the segment amounts use a consistent exposure definition, or when the components are reported separately by basis or classification.

E_j=N_j\frac{d_j}{D}

This time-proration formula applies when $N_j$ is a constant exposure count throughout segment $j$, exposure is earned uniformly by day, and $d_j$ of $D$ days are covered. It does not apply automatically to changing payroll, sales, or other exposure that varies within the segment. For those bases, use actual earned amounts or a documented allocation method appropriate to the data.

Why Segmentation Preserves the Exposure Measure

Segmentation is not merely an administrative convention. It follows from adding exposure over nonoverlapping periods. If $n(t)$ denotes the number of insured units at time $t$, then unit-time exposure over an interval is the integral of that count over time. Splitting the interval at a change date does not alter the total; it allows each portion to use the correct count.

\int_{a}^{c} n(t)\,dt = \int_{a}^{b} n(t)\,dt+\int_{b}^{c} n(t)\,dt

The equality holds because the intervals from $a$ to $b$ and from $b$ to $c$ are adjacent and do not overlap. If the insured-unit count is constant within each segment, each integral reduces to the count multiplied by the segment duration. A midterm change therefore calls for a new segment, not for applying the changed count retroactively.

Loss attribution uses a related but distinct rule. Exposure is accumulated over time; a loss is assigned using its occurrence date and the coverage and risk characteristics then in force. The date a claim is reported, reserved, or paid may be later than the occurrence date. Those later dates may matter for valuation and maturity, but they do not ordinarily move the event into a different operational segment.

Worked Example: A Classification Change and Actual Payroll

A hypothetical workers compensation policy runs for calendar year 2025. A covered operation is in Classification A through April 30; a new endorsement makes Classification B effective May 1. Audited payroll is $240,000 for Classification A and $490,000 for Classification B. Covered losses, stated on the same valuation basis, are $18,000 from an event on March 15, $5,000 from an event on April 30, $12,000 from an event on May 1, and $27,000 from an event on November 15. Determine the exposure and loss totals by classification. Interest, mortality, and a probability distribution are not assumptions in this deterministic accounting exercise.

Method. Use the actual payroll by classification, and use each loss's occurrence date to identify the classification in force. The endorsement's effective date is May 1, so April 30 remains in the old segment and May 1 is in the new one.

Exposure. Classification A exposure is the audited amount of $240,000. Classification B exposure is $490,000. The total is $240,000 + $490,000 = $730,000. No time-proration is needed because payroll by segment is supplied directly.

Losses in A. Both March 15 and April 30 precede May 1. Thus A losses are $18,000 + $5,000 = $23,000.

Losses in B. May 1 is the effective date of the endorsement, and November 15 is also after that date. Thus B losses are $12,000 + $27,000 = $39,000.

Check. The class totals reconcile: $23,000 + $39,000 = $62,000, the sum of all four losses. The common mistake is to assign the April 30 loss to B because the endorsement was made during the policy term; the correction is to follow the effective date, not the policy term as a whole.

Worked Example: Vehicle Changes Followed by Cancellation

A hypothetical fleet policy covers five vehicles from January 1 through March 31, 2025. Two vehicles are added effective April 1, so seven are covered through June 30. Effective July 1, three vehicles are removed, leaving four covered through September 30. The policy is cancelled effective October 1, with no coverage beginning that day. Calculate vehicle-year exposure using a 365-day year, and identify which listed losses belong in the experience. Covered losses occurred on February 10 ($8,000), May 12 ($14,000), August 20 ($20,000), September 30 ($6,000), and October 3 ($9,000). Interest, mortality, and a probability distribution are not assumed.

Method. Calculate vehicle-days within each distinct covered segment, add them, and divide by 365. A loss is eligible for this experience only if its occurrence date falls within the stated coverage period.

January 1–March 31. In 2025 this is 31 + 28 + 31 = 90 days. Exposure is 5 × 90 = 450 vehicle-days.

April 1–June 30. This is 30 + 31 + 30 = 91 days. Exposure is 7 × 91 = 637 vehicle-days.

July 1–September 30. This is 31 + 31 + 30 = 92 days. Exposure is 4 × 92 = 368 vehicle-days.

Total exposure is 450 + 637 + 368 = 1,455 vehicle-days. Dividing by 365 gives 1,455 / 365 = 3.9863 vehicle-years, rounded to four decimal places.

The February, May, August, and September 30 losses occurred during the stated covered intervals, so their total is $8,000 + $14,000 + $20,000 + $6,000 = $48,000. The October 3 loss is after cancellation took effect and is excluded from this policy's experience. The September 30 loss is included because coverage extends through that date. A common error is to count exposure through December 31 because that was the scheduled expiration; the correction is to stop earning exposure at the cancellation's effective date, following the stated coverage boundary.

Worked Example: A Changed Operation and a Late-Reported Claim

A hypothetical distributor's policy covers warehouse and delivery operations throughout 2025. Effective July 1, delivery operations cease and the insured retains only its warehouse operation. Actual payroll is $210,000 for warehouse work and $180,000 for delivery work from January 1 through June 30; from July 1 through December 31 it is $240,000 for warehouse work and zero for delivery work. A $7,000 driver injury occurred June 28 but was reported August 10. Other covered losses are a $10,000 driver injury on March 29 and a $12,000 warehouse injury on July 18. Find payroll exposure and losses by operation. Interest, mortality, and a probability distribution are not assumed.

Method. Keep actual payroll in its corresponding operation, and assign each loss by the operation and coverage in force on its occurrence date. Reporting after the change does not change when the event occurred.

Warehouse payroll. First-half payroll of $210,000 plus second-half payroll of $240,000 gives $450,000.

Delivery payroll. First-half payroll of $180,000 plus second-half payroll of zero gives $180,000. Total payroll is $450,000 + $180,000 = $630,000.

Losses. The March 29 driver injury and June 28 driver injury both occurred while delivery operations were covered. Delivery losses are $10,000 + $7,000 = $17,000. The July 18 warehouse injury occurred after the change, so warehouse losses are $12,000.

Check. Losses by operation total $17,000 + $12,000 = $29,000, matching the sum of the three listed claims. The $7,000 claim remains a delivery loss even though it was reported after delivery ceased; the common mistake is to classify claims by report date rather than occurrence date.

Practical Controls and Common Traps

A reliable experience record should make it possible to reproduce both the exposure calculation and the loss assignment. Maintain a dated change log that identifies the endorsement or cancellation, its effective date, affected units or operations, and the exposure basis used. Reconcile the segments to policy records and the applicable exposure audit. For losses, retain occurrence dates, covered locations or operations, and any information needed to confirm the coverage terms in force.

Do not annualize a partial term as if it were a full term. Earn only the covered time or the actual exposure incurred.
Do not apply a new classification or operation to earlier dates. Use its effective date to establish the segment boundary.
Do not use report date as a substitute for occurrence date. A delayed report does not ordinarily move an event into a later segment.
Do not assume all changes affect every exposure basis. A vehicle endorsement may change vehicle count, while a separate payroll basis may require actual payroll records.
Do not include time after cancellation. Follow the effective date and coverage boundary shown in the policy documents.

Where the available data do not support precise allocation, document the method and its limitations rather than implying a level of accuracy the records cannot provide. Any method should be applied consistently to comparable risks and should preserve a clear audit trail. These controls are aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus, which emphasizes defensible experience-period design.

Key takeaway: Split exposure at each material effective date, measure every segment on the appropriate and consistent basis, and assign each loss using its occurrence date and the coverage and operation then in force. Reconcile the pieces before using the experience in a rating plan.

Check Your Understanding

For each problem, use the stated effective dates and exposure basis. Assume the listed losses are otherwise covered unless the facts say otherwise.

A policy covers 12 identical insured units for the first 90 days of 2025 and 10 units for the next 91 days. Using a 365-day year, calculate exposure in unit-years. (Answer: 1,990 / 365 = 5.4521 unit-years, rounded to four decimal places.)
An endorsement changes a risk from Classification A to Classification B effective August 1. A covered loss of $4,500 occurs July 31, and another covered loss of $6,000 occurs August 1. To which classification does each loss belong, and what are the loss totals by classification? (Answer: A has $4,500; B has $6,000.)
A policy covers 20 locations from January 1 through November 30, 2025, and is cancelled effective December 1. Using a 365-day year and uniform location exposure, calculate location-year exposure. Does a covered November 30 loss belong in the policy experience? (Answer: 20 × 334 / 365 = 18.3014 location-years; yes, the November 30 loss is within the covered period.)
A business has $150,000 payroll in Classification A through March 31, then changes to Classification B effective April 1 and incurs $300,000 payroll in B for the rest of the year. A $10,000 loss occurs March 30, a $7,000 loss occurs March 29 but is reported May 10, and a $12,000 loss occurs April 2. State exposure and loss totals by classification. (Answer: A exposure $150,000 and losses $17,000; B exposure $300,000 and losses $12,000.)
A fleet has three vehicles from January 1 through March 31, four from April 1 through June 30, and no coverage beginning July 1 because the policy is cancelled effective that date. Using 2025 and a 365-day year, calculate vehicle-year exposure and determine whether losses occurring June 30 and July 1 are included. (Answer: (3 × 90 + 4 × 91) / 365 = 634 / 365 = 1.7370 vehicle-years, rounded to four decimal places; June 30 is within coverage, July 1 is not.)

What You'll Learn