Why Unequal Accident Periods Need Exposure Adjustment
Accident periods rarely represent equal volumes of risk. One year may contain substantially more earned policies, insured vehicles, or payroll than another. Comparing total losses alone can therefore make a high-volume period look worse even when its loss cost per unit of exposure is lower. The useful comparison is usually loss relative to an exposure measure that matches the losses being analyzed.
This tutorial focuses on calculating and interpreting that exposure-adjusted experience. It also shows why the exposure unit must be consistent across periods: a policy-year and a policy-month do not represent the same quantity. These calculations can inform a later projection, but an exposure-adjusted loss cost by itself is not a completed reserve estimate. In particular, it does not automatically account for claim development, changes in coverage, or other adjustments that a reserve analysis may require.
The choice of exposure measure depends on the business and the purpose of the analysis. Examples include earned policy-years, vehicle-years, or payroll in thousands of dollars. The numerator and denominator must refer to the same underlying risks and time basis. For example, dividing losses from a group of earned policies by a premium exposure from a different group would not create a meaningful loss cost merely because the arithmetic is possible.
Why the Pooled Rate Is Exposure-Weighted
A pooled loss cost is not generally the arithmetic mean of the accident-period loss costs. Each period's rate must contribute in proportion to its exposure volume. This follows directly from the definition of a pooled rate.
The identity holds because \(L_i=E_i r_i\) for each period. Substituting these equalities into total losses gives \(\sum_i L_i=\sum_i E_i r_i\); dividing by total exposure yields the weighted-average expression. The weights are nonnegative and sum to one, so the pooled rate lies between the smallest and largest period rates when all exposures are positive.
A simple average assigns the same weight to every accident period regardless of volume. It answers a different question: the average of the period rates when each period is given equal influence. It is appropriate only when equal period weighting is intended, not when estimating the loss cost for the combined exposure.
Worked Example: Pooling Accident Periods With Different Volumes
A hypothetical insurer has the following losses and earned exposures on a consistent basis. Calculate each period's loss cost and the pooled loss cost. Compare the result with the simple average of the period rates.
| Accident period | Losses | Earned exposure |
|---|---|---|
| Period 1 | $720,000 | 1,000 policy-years |
| Period 2 | $420,000 | 500 policy-years |
| Period 3 | $900,000 | 1,500 policy-years |
Step 1: Calculate the period rates. Period 1 has \(720{,}000/1{,}000=\$720\) per policy-year. Period 2 has \(420{,}000/500=\$840\) per policy-year. Period 3 has \(900{,}000/1{,}500=\$600\) per policy-year.
Step 2: Pool the losses and exposure. Total losses are \(720{,}000+420{,}000+900{,}000=\$2{,}040{,}000\). Total exposure is \(1{,}000+500+1{,}500=3{,}000\) policy-years.
Step 3: Divide total losses by total exposure. The pooled loss cost is \(\$2{,}040{,}000/3{,}000=\$680\) per policy-year.
Step 4: Compare with the simple average. The simple average is \((720+840+600)/3=\$720\) per policy-year. It exceeds the pooled rate because it gives the 500-policy-year period the same weight as the 1,500-policy-year period. The correct exposure weights are \(1{,}000/3{,}000=1/3\), \(500/3{,}000=1/6\), and \(1{,}500/3{,}000=1/2\); their weighted rate is \(720/3+840/6+600/2=680\).
Check. Multiplying the pooled rate by the total exposure gives \(\$680(3{,}000)=\$2{,}040{,}000\), which equals the stated total losses. The method avoids the common mistake of averaging period rates equally when the question asks for the combined loss cost.
Exposure Units Must Be Consistent
A rate always has units. A loss cost of $700 per policy-year is not numerically the same rate as a loss cost per policy-month. Since one policy-year equals 12 policy-months, the same losses divided by policy-month exposure produce a rate one-twelfth as large. Converting the unit changes the numerical rate, not the underlying experience.
If every exposure is converted by the same factor \(c\), the exposure weights remain unchanged: \(cE_i/\sum_j(cE_j)=E_i/\sum_jE_j\). The pooled rate expressed per converted unit changes inversely by that factor: \(\sum_iL_i/\sum_i(cE_i)=r_{\text{pool}}/c\). For example, expressing exposure in policy-months rather than policy-years multiplies the denominator by 12 and divides the numerical loss cost by 12. If only some periods are converted, however, the relative weights are wrong until all periods are put on one basis.
Worked Example: Reconciling Policy-Years and Policy-Months
A hypothetical portfolio has $840,000 of losses over 1,200 policy-years in one accident period. A second period has $1,200,000 of losses over 18,000 policy-months. Find the pooled loss cost per policy-year and per policy-month.
Step 1: Convert the second exposure to policy-years. Since 12 policy-months equal one policy-year, \(18{,}000/12=1{,}500\) policy-years.
Step 2: Pool on the policy-year basis. Total losses are \(\$840{,}000+\$1{,}200{,}000=\$2{,}040{,}000\). Total exposure is \(1{,}200+1{,}500=2{,}700\) policy-years. Therefore, the pooled rate is \(\$2{,}040{,}000/2{,}700=\$755.555\ldots\), or \(\$755.56\) per policy-year.
Step 3: Express the rate per policy-month. Divide by 12: \(\$755.555\ldots/12=\$62.9629\ldots\), or \(\$62.96\) per policy-month.
Step 4: Check using policy-months throughout. The first exposure is \(1{,}200(12)=14{,}400\) policy-months. Total exposure is \(14{,}400+18{,}000=32{,}400\) policy-months. Then \(\$2{,}040{,}000/32{,}400=\$62.9629\ldots\) per policy-month, agreeing with Step 3.
Common mistake and correction. Dividing $2,040,000 by \(1{,}200+18{,}000=19{,}200\) treats policy-years and policy-months as interchangeable. That denominator has no consistent exposure unit and gives a meaningless numerical rate. Convert one exposure basis to the other before pooling. The corrected results are $755.56 per policy-year and $62.96 per policy-month, with rounding to the nearest cent.
Adjustments Before Pooling
Accident periods may also require an adjustment to put losses on a common level before they are pooled. For example, a problem may specify factors to bring historical losses to a stated cost level. Such factors are applied to the losses as directed, while exposure remains on its defined basis unless the problem specifies an exposure adjustment too. After adjustment, each numerator and denominator must still match in scope and units.
The adjustment is not a license to alter data without an assumption. State what is being adjusted, which factor applies to each period, and what common level the adjusted amounts represent. If factors reflect a change in claim costs, they do not automatically correct for a shift in risk mix, coverage, reporting, or exposure measurement. The calculation is only as interpretable as those assumptions.
Worked Example: Adjusting Losses to a Common Cost Level
A hypothetical insurer is asked to calculate an exposure-weighted loss cost at the current cost level. The problem supplies a factor for each period. Assume the stated losses and exposures are otherwise comparable and that the factors apply only to losses.
| Accident period | Losses | Exposure | Loss adjustment factor |
|---|---|---|---|
| Period 1 | $480,000 | 800 policy-years | 1.10 |
| Period 2 | $840,000 | 1,200 policy-years | 1.05 |
| Period 3 | $1,600,000 | 2,000 policy-years | 1.00 |
Step 1: Adjust each loss amount. Period 1 becomes \(\$480{,}000(1.10)=\$528{,}000\). Period 2 becomes \(\$840{,}000(1.05)=\$882{,}000\). Period 3 remains \(\$1{,}600{,}000(1.00)=\$1{,}600{,}000\).
Step 2: Sum adjusted losses and exposure. Adjusted losses total \(\$528{,}000+\$882{,}000+\$1{,}600{,}000=\$3{,}010{,}000\). Exposure totals \(800+1{,}200+2{,}000=4{,}000\) policy-years.
Step 3: Calculate the adjusted pooled loss cost. The result is \(\$3{,}010{,}000/4{,}000=\$752.50\) per policy-year.
Step 4: Use the rate for a stated exposure projection. If, under the same assumptions, a hypothetical future period is expected to have 2,500 policy-years, multiplying gives \(2{,}500(\$752.50)=\$1{,}881{,}250\) of projected losses at that cost level.
Check. The adjusted period rates are $660, $735, and $800 per policy-year. Their exposure weights are \(800/4{,}000=0.20\), \(1{,}200/4{,}000=0.30\), and \(2{,}000/4{,}000=0.50\). Thus \(0.20(660)+0.30(735)+0.50(800)=132+220.50+400=\$752.50\). The projection is an exposure-based illustration, not a reserve estimate; it does not include any separate development or reserve assumptions.
Interpreting Experience for Projections
An exposure-adjusted rate helps separate volume from cost. If a period has twice the exposure of another and the same loss cost, its expected total losses under the same conditions would be twice as large. This is why a rate may be applied to a future exposure estimate when the task calls for a basic experience projection.
That projection requires more than arithmetic. The experience must be relevant to the projected risks; the loss basis and exposure definition must be consistent; and any assumed changes in cost level, coverage, or mix must be handled explicitly. A projection based on policy-years is not compatible with a forecast expressed in policy-months until one is converted. Nor should an exposure-weighted average of observed losses be mistaken for an ultimate-loss estimate when losses are immature or otherwise incomplete.
Exposure consistency also matters when experience feeds later credibility or reserving calculations. A period with a larger true exposure should generally contribute more to a pooled experience rate because it represents more risk units. But a unit error can artificially magnify or shrink a period's apparent volume, changing its contribution even though the underlying claims have not changed. This tutorial establishes the exposure calculation; it does not specify a credibility formula or a reserve method.
Common Exam Traps
- Using a simple average instead of a pooled rate. If the target is total losses per total exposure, calculate \(\sum L_i/\sum E_i\), or use the equivalent exposure-weighted average.
- Mixing exposure units. Convert all periods to policy-years, policy-months, or another single stated unit before summing or comparing rates.
- Assuming a unit conversion leaves the rate's numerical value unchanged. Converting every exposure from policy-years to policy-months leaves the weights unchanged but divides the rate per unit by 12.
- Applying a loss-level factor to exposure without instruction. If the stated factors adjust losses only, adjust the numerators and retain the original compatible exposure denominators.
- Calling a basic projected loss amount a reserve. Multiplying a loss cost by future exposure does not, by itself, estimate unpaid losses or account for development.
Check Your Understanding
For each problem, use the stated loss basis and exposure units. Round monetary rates to the nearest cent where needed.
- A hypothetical account has losses of $390,000 over 600 policy-years in one period and $910,000 over 1,400 policy-years in another. Calculate the pooled loss cost per policy-year. (Answer: $650.00 per policy-year)
- Two periods have losses of $360,000 over 600 policy-years and $540,000 over 900 policy-years. What is the exposure-weighted pooled rate, and why would an equal-period average not change it in this case? (Answer: $600.00 per policy-year; both period rates are $600.00)
- A portfolio has $620,000 of losses over 1,000 policy-years and $1,500,000 over 24,000 policy-months. Find the pooled rate per policy-year and per policy-month. (Answer: $706.67 per policy-year and $58.89 per policy-month)
- A problem specifies loss adjustment factors of 1.20 and 1.10 for two periods. Losses are $275,000 over 500 policy-years and $650,000 over 1,000 policy-years, respectively. After applying the factors to losses only, calculate the pooled loss cost. If future exposure is 1,800 policy-years, calculate the corresponding basic projected losses. (Answer: $696.67 per policy-year; $1,254,000)
- One period has $520,000 in losses over 800 policy-years. Another has $780,000 over 12,000 policy-months. A candidate divides total losses by \(800+12{,}000\). Identify the error and give the correct pooled rate per policy-year. (Answer: The denominator mixes units; convert 12,000 policy-months to 1,000 policy-years, giving $722.22 per policy-year)