Reconciling Component Trends With Aggregate Trend

Why Component Trends Need Reconciliation

A pure premium is the expected loss per exposure. It can be expressed as claim frequency multiplied by claim severity, so it is natural to expect a pure premium trend to equal the product of a frequency trend and a severity trend. That relationship is exact when all three trends use compatible definitions and are calculated from the same aggregate portfolio at the same endpoints.

In practice, frequency and severity may be estimated separately, with different weights or models. Segment mix can change, and an estimate based on weighted component trends need not equal the directly observed aggregate trend. The key is to identify what each factor measures, then reconcile the factors using consistent weights and definitions.

Definition: For a period with $E$ exposures, $N$ claims, and total loss $L$, claim frequency is $F=N/E$, claim severity is $S=L/N$ when $N>0$, and pure premium is $P=L/E$. Therefore, $P=F S$. A trend factor from period 0 to period 1 is the period 1 measure divided by the period 0 measure.

The identity $P=FS$ is an accounting relationship: multiplying claims per exposure by loss per claim gives loss per exposure. It does not depend on a statistical model. For a trend comparison, however, the measures must refer to compatible claim counts, losses, exposure bases, periods, and treatments of claims. If severity is calculated using a different claim definition or loss basis than the one used in the pure premium, the identity will not reconcile the estimates.

When the Product Relationship Is Exact

Let $F_t$, $S_t$, and $P_t$ be frequency, severity, and pure premium for the same portfolio in period $t$, where $t$ is 0 or 1. Define their trend factors as $T_F=F_1/F_0$, $T_S=S_1/S_0$, and $T_P=P_1/P_0$. If $P_t=F_tS_t$ in both periods, then:

T_P =\frac{P_1}{P_0} =\frac{F_1S_1}{F_0S_0} =\left(\frac{F_1}{F_0}\right)\left(\frac{S_1}{S_0}\right) =T_FT_S.

This shows why the product rule is exact for pooled measures. The formula is not a general rule for multiplying any two separately estimated trend factors. In particular, an arithmetic average of segment frequency trends and an arithmetic average of segment severity trends usually do not equal the aggregate pure premium trend when segment weights differ or change.

Worked Example: Reconciling Pooled Frequency and Severity

A hypothetical insurer has 2,000 exposures in each of two periods. In period 0, it records 400 claims and $600,000 of losses. In period 1, it records 484 claims and $798,600 of losses. Calculate the frequency, severity, and pure premium trend factors and verify the reconciliation. The common mistake to avoid is comparing claim-count growth directly with loss-cost growth without allowing for the exposure base.

Period 0 frequency: $F_0=400/2{,}000=0.20$ claims per exposure. Period 1 frequency: $F_1=484/2{,}000=0.242$. Thus $T_F=0.242/0.20=1.21$, a 21% increase.

Period 0 severity: $S_0=\$600{,}000/400=\$1{,}500$ per claim. Period 1 severity: $S_1=\$798{,}600/484=\$1{,}650$ per claim. Thus $T_S=\$1{,}650/\$1{,}500=1.10$, a 10% increase.

Period 0 pure premium: $P_0=\$600{,}000/2{,}000=\$300$ per exposure. Period 1 pure premium: $P_1=\$798{,}600/2{,}000=\$399.30$. Therefore $T_P=\$399.30/\$300=1.331$.

Check the product: $T_FT_S=1.21(1.10)=1.331$, matching the directly calculated pure premium factor. The calculation is consistent because both component measures use the same claims, losses, exposure base, and periods.

Why Segment-Weighted Components May Not Reconcile

For a portfolio divided into segments $i$, aggregate frequency is an exposure-weighted average of segment frequencies. If $E_{i,t}$ is segment exposure and $F_{i,t}$ is its frequency, then:

F_t=\frac{\sum_i E_{i,t}F_{i,t}}{\sum_i E_{i,t}}.

Aggregate severity is instead a claim-weighted average of segment severities. With $N_{i,t}$ claims and severity $S_{i,t}$:

S_t=\frac{\sum_i N_{i,t}S_{i,t}}{\sum_i N_{i,t}}.

These weights are different: frequency uses exposure, while severity uses claims. Moreover, claim weights can change as segment frequencies change. As a result, an average of segment severity trend factors using period 0 claim weights is generally not the trend in pooled severity. The exact pooled severity trend uses the pooled severity in each period, including each period’s actual claim mix.

Worked Example: Changing Claim Mix Alters Pooled Severity

Consider a hypothetical portfolio with two segments, each having 1,000 exposures in both periods. Segment A has frequency 0.10 and severity $1,000 in period 0; its period 1 frequency is 0.11 and its severity is $1,200. Segment B has frequency 0.20 and severity $2,000 in both periods. Calculate the aggregate pure premium trend and compare it with a severity trend formed by weighting segment severity trends using period 0 claim counts. The common mistake is to treat those old claim weights as unchanged when frequency shifts the claim mix.

Period 0 totals. Segment A has $1{,}000(0.10)=100$ claims and $100,000 of losses. Segment B has $1{,}000(0.20)=200$ claims and $400,000 of losses. In total, there are 300 claims and $500,000 of losses over 2,000 exposures. Thus $F_0=300/2{,}000=0.15$, $S_0=\$500{,}000/300=\$1{,}666.67$, and $P_0=\$500{,}000/2{,}000=\$250$.

Period 1 totals. Segment A has $1{,}000(0.11)=110$ claims and $110(\$1{,}200)=\$132{,}000$ of losses. Segment B remains at 200 claims and $400,000 of losses. In total, there are 310 claims and $532,000 of losses over 2,000 exposures. Thus $F_1=310/2{,}000=0.155$, $S_1=\$532{,}000/310=\$1{,}716.13$, and $P_1=\$532{,}000/2{,}000=\$266$.

The aggregate factors are $T_F=0.155/0.15=1.033333$, $T_S=(532{,}000/310)/(500{,}000/300)=1.029677$, and $T_P=266/250=1.064$. Their pooled relationship checks: $1.033333(1.029677)$ is approximately $1.064$. Factors are shown to six decimals for checking; the underlying ratios reconcile exactly.

Now weight the segment severity factors using period 0 claim counts. Segment A’s severity factor is $1.20$, and segment B’s is $1.00$. The period 0 claim weights are $100/300=1/3$ for A and $200/300=2/3$ for B. Their weighted average is $(1/3)(1.20)+(2/3)(1.00)=1.066667$. That is not the pooled severity factor of $1.029677$. Applying the old-weight average to the pooled frequency factor gives $1.033333(1.066667)=1.102222$, not the aggregate pure premium factor of $1.064$.

The difference is not a failure of $P=FS$. The period 1 pooled severity correctly uses the new claim weights $110/310$ for A and $200/310$ for B. The comparison used period 0 weights instead. The aggregate calculation avoids that mismatch by pooling each period’s actual claims and losses before calculating its severity.

Reconcile Segment Pure Premium Trends With Loss Weights

A second useful reconciliation starts directly from segment pure premiums. If exposures are unchanged by segment between the two periods, the aggregate pure premium trend is the baseline-loss-weighted average of the segment pure premium trend factors. The baseline loss weight for segment $i$ is its period 0 loss divided by total period 0 loss. This follows because projected aggregate losses are the sum of baseline segment losses multiplied by their respective trend factors.

T_P =\frac{\sum_i L_{i,0}T_{P,i}}{\sum_i L_{i,0}} =\sum_i \left(\frac{L_{i,0}}{\sum_j L_{j,0}}\right)T_{P,i}.

Here $T_{P,i}$ is the segment pure premium trend factor. If segment exposures change, or if the intended aggregate measure uses a different exposure distribution, the calculation must use the appropriate projected segment exposures as well. Do not apply the fixed-exposure formula without checking that condition.

Worked Example: Weighting Segment Pure Premium Trends

A hypothetical portfolio has two segments with exposures held constant. Segment A has 600 exposures and a baseline pure premium of $500; its frequency and severity trend factors are 1.08 and 1.05. Segment B has 400 exposures and a baseline pure premium of $1,500; its corresponding factors are 0.95 and 1.10. Calculate the segment pure premium factors and the aggregate factor. The common mistake is to take an unweighted average of the segment factors.

Segment A. Its pure premium factor is $1.08(1.05)=1.134$. Its baseline loss is $600(\$500)=\$300{,}000$.

Segment B. Its pure premium factor is $0.95(1.10)=1.045$. Its baseline loss is $400(\$1{,}500)=\$600{,}000$.

Baseline aggregate loss is $\$300{,}000+\$600{,}000=\$900{,}000$, and total exposure is 1,000, so baseline aggregate pure premium is $900. Projected losses are $\$300{,}000(1.134)=\$340{,}200$ for A and $\$600{,}000(1.045)=\$627{,}000$ for B. Total projected loss is $967,200, giving projected aggregate pure premium of $967.20 and aggregate trend factor $\$967.20/\$900=1.074667$.

The baseline loss weights are $300{,}000/900{,}000=1/3$ for A and $600{,}000/900{,}000=2/3$ for B. The weighted factor is $(1/3)(1.134)+(2/3)(1.045)=1.074667$, which agrees. The simple average, $(1.134+1.045)/2=1.0895$, is incorrect because the two segments do not contribute equally to baseline loss.

Practical Checks and Common Traps

When two component estimates do not multiply to the aggregate estimate, first check the calculation basis rather than assuming one estimate is wrong. A reconciliation should make explicit which portfolio, periods, claim count, loss measure, and exposure measure are being compared.

Mixing weights: frequency is exposure-weighted, while severity is claim-weighted. A segment’s share of exposure is not generally its share of claims or losses.
Holding weights fixed without justification: a baseline claim-weighted severity trend may not reflect a changed claim mix in the later period.
Comparing different measures: a frequency factor based on reported claims and a severity factor based on a differently defined loss or claim population will not necessarily reproduce the same pure premium trend.
Using an unweighted average: average segment factors are not aggregate factors unless their weights happen to match the relevant portfolio contributions.
Rounding too early: retain full precision in intermediate calculations, then round the reported factors. Small rounding differences can remain even when the definitions and weights agree.

The practical reconciliation is to calculate aggregate frequency and severity from the same period-level totals, or to project segment losses using explicit segment weights and exposures. If the separately estimated factors still differ, document the weighting or estimation choices responsible for the difference. A discrepancy is informative: it identifies a difference in composition or method that a single aggregate number can conceal.

Key takeaway: The aggregate pure premium trend equals the product of aggregate frequency and aggregate severity trends when each is calculated consistently from the same portfolio and periods. Independently weighted component estimates may not multiply to that result; reconcile them using the correct period-specific claim weights or baseline segment loss weights.

Check Your Understanding

For each hypothetical situation, calculate the requested factor and check which weights or definitions are being used.

A portfolio has 500 claims and $750,000 of losses in period 0, then 550 claims and $907,500 of losses in period 1. Exposure is 2,500 in both periods. Find the frequency, severity, and pure premium trend factors, and verify the product relationship. (Answer: Frequency $1.1000$, severity $1.1000$, pure premium $1.2100$; $1.1000 \times 1.1000=1.2100$.)
Two segments have baseline losses of $240,000 and $360,000, with pure premium trend factors 1.05 and 0.90. Exposures are constant by segment. Find the aggregate trend factor and compare it with the unweighted average. (Answer: Aggregate $0.9600$; unweighted average $0.9750$.)
In period 0, segments A and B have 80 and 120 claims, with severity trend factors 1.25 and 0.95. Using period 0 claim weights, calculate the weighted severity factor. (Answer: $(80/200)(1.25)+(120/200)(0.95)=1.0700$.)
A portfolio’s aggregate frequency factor is 1.04 and its consistently pooled severity factor is 0.98. Find the aggregate pure premium factor and percentage change. (Answer: $1.04(0.98)=1.0192$, an increase of 1.92%.)
Two constant-exposure segments have baseline losses of $100,000 and $300,000 and segment pure premium trend factors of 1.20 and 0.90. Calculate projected total loss and the aggregate trend factor. (Answer: Projected loss $=\$100{,}000(1.20)+\$300{,}000(0.90)=\$390{,}000$; baseline loss $=\$400{,}000$; aggregate factor $=0.9750$.)

What You'll Learn