Why Pool Claims?
Severity measures the average loss amount associated with a claim. When experience is recorded in several periods, the analyst may have each period’s total loss and claim count rather than every individual claim amount. Pooling combines those totals and counts into one estimate. The key is to divide total losses by total claims, not to give every period equal influence regardless of how many claims it contains.
That distinction matters whenever period claim counts differ. A period with many claims contributes more observations to an overall average than a period with only a few claims. An unweighted average of period averages instead gives each period the same influence. Both calculations can be valid summaries, but they answer different questions.
The numerator and denominator must describe matching experience. For example, if the numerator is incurred loss, it should include losses corresponding to the claims counted in the denominator, using a consistent definition of incurred loss across periods. If one period’s count includes claims excluded from its loss total, the calculated ratio is not a meaningful average severity for the combined claims.
From Period Severities to the Pooled Estimate
When every period has at least one claim, the pooled result can also be calculated from the period severities by weighting each one by its share of the total claims:
The weight for period \(j\) is its fraction of all pooled claims. The weights are nonnegative and sum to one. Consequently, the pooled severity lies between the smallest and largest period severities, provided all period counts are positive. A period with more claims has a larger weight; a period with fewer claims has a smaller weight.
Check that the loss and count refer to the same claims and use consistent definitions.
Calculate \(L_{\text{total}}=\sum_j L_j\) and \(N_{\text{total}}=\sum_j N_j\).
Calculate \(s_{\text{pool}}=L_{\text{total}}/N_{\text{total}}\), then check the units and arithmetic.
State clearly that it gives equal weight to periods, rather than to individual claims.
Why Claim Counts Determine the Weights
The relationship follows directly from the definition of a period severity. For every period with \(N_j>0\), its aggregate loss is \(L_j=N_js_j\). Substituting that identity into the pooled formula gives:
Each claim contributes its loss once to the numerator and one claim to the denominator. A period containing 60 claims therefore contributes six times as many claim observations as a period containing 10 claims. Giving their period averages equal weights would not reproduce the average over all individual claims unless the claim counts happened to be equal, or the relevant period severities were equal.
This is a descriptive identity; it does not by itself establish that experience from different periods is suitable for a particular future estimate. The intended population and the basis of the losses still matter. If the periods represent different claim populations, the pooled value remains the average for the combined observed claims, but it need not be the right estimate for any one of those populations separately.
Worked Example: Pooling Two Periods
A hypothetical equipment insurer reviews two periods. Period 1 has 40 claims and $120,000 in aggregate loss. Period 2 has 60 claims and $210,000 in aggregate loss. Calculate each period’s severity, the pooled severity, and the unweighted average of the two period severities. Explain which result is the average across all 100 claims.
Method. Use total loss divided by total claim count for the pooled estimate. Calculate period severities first to make the weighting difference visible.
Period 1 severity. Divide its aggregate loss by its count: $$ s_1=\frac{\$120{,}000}{40}=\$3{,}000. $$
Period 2 severity. $$ s_2=\frac{\$210{,}000}{60}=\$3{,}500. $$
Pooled severity. Total loss is \(\$120{,}000+\$210{,}000=\$330{,}000\). Total claims are \(40+60=100\). Thus: $$ s_{\text{pool}}=\frac{\$330{,}000}{100}=\$3{,}300. $$
Unweighted average of period severities. $$ \frac{\$3{,}000+\$3{,}500}{2}=\$3{,}250. $$ This gives each period one-half of the weight. The pooled weights are \(40/100=0.4\) and \(60/100=0.6\), so the claim-weighted calculation is \(0.4(\$3{,}000)+0.6(\$3{,}500)=\$1{,}200+\$2{,}100=\$3{,}300\).
Check. The weighted calculation agrees with total loss divided by total claims. The pooled estimate, \(\$3{,}300\), is the average across all 100 claims; the unweighted period average, \(\$3{,}250\), is a different summary. Common mistake avoided: treating two periods as equally influential when their claim counts are 40 and 60.
The Effect of Unequal Claim Counts
The difference between the two averages can be substantial when counts and severities vary together. If a small period has unusually high severity, an unweighted average can make that period appear just as influential as a much larger period. Pooling reduces the influence of that small period in proportion to its share of claims.
The unweighted average is not inherently an arithmetic error. It is the mean of the period severities when the periods themselves are the units being averaged. It simply does not answer the same question as the mean loss per claim. Before calculating, identify whether the target is an average over claims or an average over periods.
Worked Example: A Small High-Severity Period
A hypothetical specialty insurer has 5 claims with average severity of $8,000 in one period and 45 claims with average severity of $2,000 in another. Calculate aggregate losses, the pooled severity, and the unweighted average of period severities. State why the two results differ.
Method. Recover each period’s total loss by multiplying its claim count by its average severity. Then divide combined loss by combined claims.
First period total loss. \(5\times \$8{,}000=\$40{,}000\).
Second period total loss. \(45\times \$2{,}000=\$90{,}000\).
Pooled calculation. Total loss is \(\$40{,}000+\$90{,}000=\$130{,}000\). Total claims are \(5+45=50\). Therefore: $$ s_{\text{pool}}=\frac{\$130{,}000}{50}=\$2{,}600. $$
Unweighted period average. $$ \frac{\$8{,}000+\$2{,}000}{2}=\$5{,}000. $$ The claim-count weights are \(5/50=0.1\) and \(45/50=0.9\). Applying them gives \(0.1(\$8{,}000)+0.9(\$2{,}000)=\$800+\$1{,}800=\$2{,}600\).
Check. The pooled value is between the two period severities and is much closer to $2,000 because 45 of the 50 claims are in that period. Common mistake avoided: using the midpoint of the two period averages as though each represented the same number of claims.
Calculation Checks and Data Conditions
A sound pooled estimate depends on more than correct division. Verify the following before interpreting the result:
- Matching claim unit: the count should measure the same kind of claim in every period. Do not combine counts defined on different bases without first reconciling them.
- Matching loss basis: use a consistent loss measure across periods, such as incurred losses throughout or paid losses throughout. Do not combine unlike measures and describe the ratio as a single comparable severity.
- Matching scope: the losses in the numerator should correspond to the claims in the denominator. A loss total that includes claims absent from the count will distort the ratio.
- Positive pooled count: the pooled formula requires \(\sum_jN_j>0\). If a period has no claims, its period severity \(L_j/N_j\) is undefined; do not calculate it by dividing by zero. A zero-claim period with zero loss can be included in the pooled totals without changing either total.
- Intended target: pooled experience describes the combined claims observed. Whether it represents a future or specific target population depends on whether those claims are relevant to that target.
Worked Example: Three Periods With Unequal Volumes
A hypothetical small-commercial portfolio has the following period results: 12 claims with average severity of $1,250; 30 claims with average severity of $1,600; and 8 claims with average severity of $2,050. Calculate aggregate loss for each period, pooled severity, and the unweighted average of the three period severities.
Method. Convert each period average back to aggregate loss, add the totals, and divide by total claims. This preserves the claim-count weighting.
Period 1 loss. \(12\times \$1{,}250=\$15{,}000\).
Period 2 loss. \(30\times \$1{,}600=\$48{,}000\).
Period 3 loss. \(8\times \$2{,}050=\$16{,}400\).
Combined totals. Aggregate loss is \(\$15{,}000+\$48{,}000+\$16{,}400=\$79{,}400\). Claim count is \(12+30+8=50\). The pooled severity is: $$ s_{\text{pool}}=\frac{\$79{,}400}{50}=\$1{,}588. $$
Unweighted period average. The sum of the three period severities is \(\$1{,}250+\$1{,}600+\$2{,}050=\$4{,}900\), so the unweighted average is \(\$4{,}900/3=\$1{,}633.33\), rounded to the nearest cent.
Check. The claim weights are \(12/50=0.24\), \(30/50=0.60\), and \(8/50=0.16\). Their weighted result is \(0.24(\$1{,}250)+0.60(\$1{,}600)+0.16(\$2{,}050)=\$300+\$960+\$328=\$1{,}588\), matching the pooled ratio. Common mistake avoided: dividing by three periods instead of by the 50 claims when the requested measure is mean loss per claim.
Common Exam Trap: Averaging Averages
A frequent error is to calculate the mean of the period severities and label it pooled severity. For the two-period example with 5 claims at $8,000 and 45 claims at $2,000, that incorrect calculation is:
The correction is to use the underlying claim totals: \((5\times \$8{,}000+45\times \$2{,}000)/(5+45)=\$130{,}000/50=\$2{,}600\). The unweighted answer is not the average across the 50 claims because it gives the five-claim period the same weight as the 45-claim period. Before selecting a calculation, check the wording: “average of period severities” calls for equal period weights, while “average per claim across combined experience” calls for pooling losses and claims.
Check Your Understanding
For each problem, identify whether the requested result is a pooled average per claim or an unweighted average across periods.
- A hypothetical repair program records 25 claims and $37,500 of loss in Period A, and 75 claims and $150,000 of loss in Period B. Find the pooled severity and the unweighted average of the period severities. (Answer: pooled severity = $1,875; unweighted period average = $1,750.)
- A hypothetical liability book has 10 claims averaging $4,800 in one period and 40 claims averaging $3,000 in another. Find the aggregate loss and pooled severity. (Answer: aggregate loss = $168,000; pooled severity = $3,360.)
- Three hypothetical periods have claim counts of 6, 14, and 30, with average severities of $900, $1,400, and $1,700, respectively. Find the total loss and pooled severity. (Answer: total loss = $76,000; pooled severity = $1,520.)
- A hypothetical insurer reports $22,500 of loss for 15 claims in one period and $99,000 for 45 claims in another. Calculate the pooled severity and compare it with the unweighted average of the period severities. (Answer: pooled severity = $2,025; unweighted period average = $1,850.)
- A hypothetical program has one period with zero claims and zero loss, a second period with 20 claims and $70,000 of loss, and a third period with 10 claims and $20,000 of loss. Find the pooled severity and explain whether the zero-claim period has a period severity. (Answer: pooled severity = $3,000; the zero-claim period severity is undefined.)