Separating Rate-Level Effects From Policy Mix Shifts

Why Separate Rate Changes From Policy-Mix Shifts?

An insurer’s aggregate premium can change for more than one reason. It may change the rates charged within rating segments, or the portfolio may contain different proportions of policies in those segments. The total premium may also change because the number of exposures changes. These effects are distinct: a rate action changes the premium assigned to a given segment, while a mix shift changes how much weight each segment receives in the aggregate.

This distinction matters when interpreting experience and preparing a rate-level comparison. If a portfolio’s average premium rises, that increase alone does not show that rates rose: the portfolio could instead have shifted toward segments with higher premiums. Conversely, a rate increase in one segment can be partly hidden by a shift toward lower-premium segments.

Definition: Suppose a portfolio is divided into $K$ segments. Let $E_{tk}$ be the number of exposures in segment $k$ at time $t$, $E_t=\sum_{k=1}^K E_{tk}$ the total exposures, and $w_{tk}=E_{tk}/E_t$ the segment’s share of exposures. Let $r_{tk}$ be premium per exposure in segment $k$. The portfolio’s average premium per exposure is $\bar r_t=\sum_{k=1}^K w_{tk}r_{tk}$, and its aggregate premium is $P_t=E_t\bar r_t$.

For this comparison, the segment definitions, exposure measure, and premium basis must be consistent between periods. Here, a segment’s $r_{tk}$ is the premium per exposure being compared at the stated rate level; the examples treat it as given. The decomposition explains changes in the resulting premium totals. It does not, by itself, determine whether the chosen rates are adequate.

A Decomposition Requires a Stated Order

When total exposures are unchanged, the observed premium change can be split into a rate effect and a mix effect. One useful convention is to apply the rate change first at the old mix, then apply the mix change at the new rates. Define $\Delta r_k=r_{1k}-r_{0k}$ and $\Delta w_k=w_{1k}-w_{0k}$. With $E_1=E_0=E$, the effects are:

\text{Rate effect}=E\sum_{k=1}^K w_{0k}\Delta r_k, \qquad \text{Mix effect}=E\sum_{k=1}^K r_{1k}\Delta w_k.

The rate effect answers: what would the premium change be if rates changed but the old portfolio mix remained? The mix effect answers: after the rate change, what additional change results from moving to the new mix? These labels depend on the order. Reversing the order allocates the interaction between rate and mix differently, but the total change must still reconcile.

Key result: If total exposure volume is fixed, the old-mix rate effect followed by the new-rate mix effect sums exactly to the observed aggregate premium change. The decomposition is an accounting identity, not a claim that one cause occurred before the other in the real world.

Why the Components Reconcile

The identity follows by adding and subtracting the intermediate premium $E\sum_k w_{0k}r_{1k}$: the premium that would result from applying the new segment rates to the old mix. The first difference captures the rate change at the old mix; the second captures the mix change at the new rates.

\begin{aligned} P_1-P_0 &=E\sum_k w_{1k}r_{1k}-E\sum_k w_{0k}r_{0k}\\ &=E\sum_k w_{0k}(r_{1k}-r_{0k}) +E\sum_k (w_{1k}-w_{0k})r_{1k}. \end{aligned}

This derivation also shows why combining a rate effect calculated at one mix with a mix effect calculated at an unrelated rate basis can fail: the two terms must be based on the same intermediate portfolio. The order is a choice, but the intermediate state must be consistent.

Worked Example: A Uniform Rate Increase and a Mix Shift

A hypothetical insurer has 100 exposures in two segments. Initially, 70 exposures are in A and 30 are in B. Premiums per exposure are $200 in A and $500 in B. Rates then rise by 5% in both segments, while the new portfolio has 50 exposures in each segment. Separate the premium increase into a rate effect followed by a mix effect, and check the total.

Method and starting values. Use the rate-first decomposition, with the old mix for the rate effect and the new rates for the mix effect. The old weights are $w_{0A}=0.70$ and $w_{0B}=0.30$; the new weights are $w_{1A}=w_{1B}=0.50$.

Initial premium. The old average premium per exposure is $0.70(200)+0.30(500)=140+150=\$290$. Thus $P_0=100(290)=\$29{,}000$.

New segment rates and rate effect. The new rates are $r_{1A}=200(1.05)=\$210$ and $r_{1B}=500(1.05)=\$525$. The segment increases are $10 and $25. Applying them at the old mix gives $100[0.70(10)+0.30(25)]=100(7+7.50)=\$1{,}450$.

Mix effect at the new rates. At the new rates, the old mix would have average premium $0.70(210)+0.30(525)=147+157.50=\$304.50$. The new mix has average premium $0.50(210)+0.50(525)=105+262.50=\$367.50$. Therefore the mix effect is $100(367.50-304.50)=\$6{,}300$.

Reconciliation. The new aggregate premium is $100(367.50)=\$36{,}750$, an increase of $\$36{,}750-\$29{,}000=\$7{,}750$. The components give $\$1{,}450+\$6{,}300=\$7{,}750$, as required. The mix shift toward segment B contributes more to the increase than the uniform rate action.

Common mistake avoided. A mix effect of $16,800 is not consistent with these weights and new rates. The old-mix premium at the new rates is $30,450, not a value that makes the mix change $16,800. Directly comparing the two new-rate averages, $367.50 and $304.50 per exposure, gives the correct $6,300 mix effect for 100 exposures.

Symmetric Allocation of the Rate–Mix Interaction

A different convention gives equal weight to the old and new states when allocating each effect. For fixed total exposures, the symmetric rate effect uses the average of the old and new weights; the symmetric mix effect uses the average of the old and new rates:

\text{Symmetric rate effect} =E\sum_k \frac{w_{0k}+w_{1k}}{2}(r_{1k}-r_{0k}), \qquad \text{Symmetric mix effect} =E\sum_k \frac{r_{0k}+r_{1k}}{2}(w_{1k}-w_{0k}).

To see why these add to the total, consider each segment’s contribution. For one segment, write $\bar w_k=(w_{0k}+w_{1k})/2$ and $\bar r_k=(r_{0k}+r_{1k})/2$. Then $\bar w_k\Delta r_k+\bar r_k\Delta w_k=w_{1k}r_{1k}-w_{0k}r_{0k}$, by expansion and cancellation of the cross terms. Summing over segments and multiplying by $E$ gives $P_1-P_0$. The symmetric approach shares the rate–mix interaction rather than assigning it entirely to whichever component comes second.

Worked Example: Symmetric Effects When Segment Rates Move Differently

A hypothetical portfolio’s segment weights change from 0.60 and 0.40 to 0.50 and 0.50. Rates per exposure change from $300 and $700 to $330 and $650. Calculate the symmetric rate and mix effects per exposure, then reconcile them to the change in average premium.

Method and rate effect. The average weights are $(0.60+0.50)/2=0.55$ and $(0.40+0.50)/2=0.45$. The rate changes are $330-300=\$30$ and $650-700=-\$50$. Therefore the symmetric rate effect per exposure is $0.55(30)+0.45(-50)=16.50-22.50=-\$6$.

Mix effect. The average rates are $(300+330)/2=\$315$ and $(700+650)/2=\$675$. The weight changes are $0.50-0.60=-0.10$ and $0.50-0.40=0.10$. The symmetric mix effect per exposure is $315(-0.10)+675(0.10)=-31.50+67.50=\$36$.

Reconciliation. The old average premium is $0.60(300)+0.40(700)=180+280=\$460$. The new average is $0.50(330)+0.50(650)=165+325=\$490$. The observed increase is $490-460=\$30$ per exposure, and the components sum to $-6+36=\$30$. A negative rate effect is possible even when one segment’s rate rises: the rate in the other segment fell, and the symmetric calculation accounts for both.

Common mistake avoided. Do not apply the old weights to the rate changes and call the result the symmetric rate effect. That calculation would be a different, rate-first decomposition. The symmetric rate effect uses average weights, while the symmetric mix effect uses average rates.

Keep Exposure Volume Separate

If total exposures change, aggregate premium has a third source of change: volume. A transparent way to separate all three effects is to state a sequence. For example, first change rates at the old volume and mix, then change the mix at the old volume and new rates, and finally change total volume at the new average premium. This order gives:

\begin{aligned} \text{Rate effect}&=E_0\sum_k w_{0k}(r_{1k}-r_{0k}),\\ \text{Mix effect}&=E_0\sum_k r_{1k}(w_{1k}-w_{0k}),\\ \text{Volume effect}&=(E_1-E_0)\sum_k w_{1k}r_{1k}. \end{aligned}

The volume term uses the new average premium because, in this stated sequence, volume changes last. Other sequences can allocate interactions differently. Whatever sequence is selected, the intermediate states must be defined and all components must reconcile to $P_1-P_0$.

Worked Example: Separating Rate, Mix, and Volume

A hypothetical portfolio begins with 80 exposures: 60 in A and 20 in B. Rates are $100 per exposure in A and $300 in B. Later, it has 100 exposures: 50 in each segment. Both segment rates have increased by 10%. Using the sequence rate, then mix, then volume, calculate each contribution and verify the total premium change.

Initial state. The old weights are $60/80=0.75$ and $20/80=0.25$. The old average premium is $0.75(100)+0.25(300)=75+75=\$150$, so $P_0=80(150)=\$12{,}000$.

Rate effect at old volume and mix. The new rates are $100(1.10)=\$110$ and $300(1.10)=\$330$. The rate effect is $80[0.75(10)+0.25(30)]=80(7.50+7.50)=\$1{,}200$.

Mix effect at old volume and new rates. The new weights are $0.50$ and $0.50$. At the new rates, the old-mix average is $0.75(110)+0.25(330)=82.50+82.50=\$165$; the new-mix average is $0.50(110)+0.50(330)=55+165=\$220$. Thus the mix effect is $80(220-165)=\$4{,}400$.

Volume effect at the new average. There are $100-80=20$ additional exposures. The volume effect is $20(220)=\$4{,}400$.

Reconciliation. The new premium is $100(220)=\$22{,}000$. The increase is $\$22{,}000-\$12{,}000=\$10{,}000$, equal to $\$1{,}200+\$4{,}400+\$4{,}400=\$10{,}000$. This check confirms that the three effects cover the entire change without double-counting.

Common mistake avoided. Comparing aggregate totals alone would show a $10,000 increase but would not identify its sources. Also, treating all of that increase as a rate effect would wrongly attribute the mix shift and the additional 20 exposures to the rate action.

Exam Traps and Practical Checks

The decomposition is useful only when its basis is clear. Before calculating, confirm that each period uses the same segment definitions and exposure unit, identify whether the requested result is per exposure or in aggregate dollars, and state which decomposition order is being used. If the total exposure counts differ, do not hide the volume change inside the mix effect.

Mix weights must sum to one. Each period’s weights are segment exposures divided by that period’s total exposures.
Rate and mix bases must match. A rate-first calculation uses old weights for the rate effect and new rates for the mix effect; the symmetric method uses average weights and average rates.
Check the aggregate total. The sum of effects must equal $P_1-P_0$, or the average-premium change when working per exposure.
Do not infer a rate change from an average-premium change alone. Segment weights may have changed even if no segment rate changed.
State the order. Different valid orders can report different individual components while still reconciling to the same total.

Key takeaway: Rate-level effects change the premium within segments; mix effects change the segments’ weights in the portfolio; volume effects change the number of exposures. Define the basis and decomposition order, then verify that the components add to the observed premium change.

Check Your Understanding

For each problem, identify the requested decomposition and check the component totals against the aggregate change.

A hypothetical portfolio has 200 exposures. The old mix is 80% in A and 20% in B, with rates of $150 and $450. Rates rise by 4% in both segments, and the new mix is 65% in A and 35% in B. Using rate first and then mix, find the two effects and total premium increase. (Answer: Rate effect $1,680; mix effect $9,360; total increase $11,040.)
Weights shift from 0.25 and 0.75 to 0.40 and 0.60. Segment rates change from $400 and $600 to $440 and $570. Calculate the symmetric rate and mix effects per exposure and reconcile them. (Answer: Rate effect −$7.25; mix effect −$24.75; total change −$32.00 per exposure.)
A portfolio starts with 120 exposures, two-thirds in A and one-third in B. Rates are $90 and $240. Later, there are 150 exposures, 40% in A and 60% in B, with rates unchanged. Using mix first and volume second, calculate the effects and total premium change. (Answer: Mix effect $4,800; volume effect $5,400; total increase $10,200.)
A portfolio’s weights change from 0.50 in A and 0.50 in B to 0.30 in A and 0.70 in B. Rates rise by 10% from $200 and $400. Using rate first and then mix, calculate the effects per exposure and verify the change. (Answer: Rate effect $30; mix effect $44; total increase $74 per exposure.)

What You'll Learn