Separating Rate Effects From Portfolio Mix Changes

Why an Aggregate Change Needs to Be Decomposed

An average premium can rise even when no segment’s premium rate changes: the portfolio may simply contain a larger share of higher-premium risks. Conversely, a rate increase can be partly hidden if the portfolio shifts toward lower-premium segments. The same issue arises when comparing loss per exposure across periods. A raw aggregate comparison combines changes in segment-level amounts with changes in the composition of the portfolio.

A fixed-mix comparison answers a focused question: what would the portfolio average have been if segment-level rates had changed but the portfolio’s segment shares had remained fixed? This is useful for monitoring a rate-level change separately from a change in the business being written or retained. It is an arithmetic attribution, not by itself proof that a rate change caused a particular outcome.

Definition: For period $t$ and segment $j$, let $E_{tj}$ be the relevant exposure, $w_{tj}=E_{tj}/\sum_k E_{tk}$ its share of total exposure, and $r_{tj}$ the premium or loss amount per exposure. The portfolio average is $\bar r_t=\sum_j w_{tj}r_{tj}$, where the weights sum to 1. A fixed-mix comparison holds one period’s weights constant while substituting the other period’s segment-level amounts.

For premium comparisons, $r_{tj}$ can be average premium per policy or per earned exposure, provided the unit and premium basis are consistent across periods. For loss comparisons, it can be loss per exposure, using a consistent loss basis. The exposure definition, time period, and treatment of transactions must be comparable. If the segment definitions or exposure units change, a numerical decomposition can still be calculated, but its interpretation may not be meaningful without reconciliation.

Exact Rate-and-Mix Decompositions

Use subscripts 0 and 1 for the earlier and later periods. One exact decomposition first measures the segment-level rate change at the earlier period’s mix, then measures the mix change using the later period’s segment-level rates.

\bar r_1-\bar r_0 = \underbrace{\sum_j w_{0j}(r_{1j}-r_{0j})}_{\text{rate effect at the earlier mix}} + \underbrace{\sum_j (w_{1j}-w_{0j})r_{1j}}_{\text{mix effect at the later rates}}.

The first term is the change that would result from updating each segment’s amount while holding the portfolio at its earlier mix. The second term captures the effect of shifting the weights, valued at the later segment amounts. The allocation depends on this chosen order: the interaction between changing rates and changing weights is assigned to the second term.

A different, equally exact sequence applies the mix change first at the earlier rates and then measures the rate change at the later mix:

\bar r_1-\bar r_0 = \sum_j (w_{1j}-w_{0j})r_{0j} + \sum_j w_{1j}(r_{1j}-r_{0j}).

If an analysis needs a decomposition that does not privilege either order, the two effects can be averaged. This symmetric allocation splits the rate-and-mix interaction equally between the effects.

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

Derivation: Start with $\bar r_1-\bar r_0=\sum_j(w_{1j}r_{1j}-w_{0j}r_{0j})$. Add and subtract $w_{0j}r_{1j}$ within each segment. Then $w_{1j}r_{1j}-w_{0j}r_{0j}=w_{0j}(r_{1j}-r_{0j})+(w_{1j}-w_{0j})r_{1j}$. Summing over segments gives the first decomposition exactly. No approximation is involved.

Worked Example: Separating Premium Rate and Mix Changes

A hypothetical insurer compares average annual premium per policy in two segments. Segment A has lower premiums than Segment B. In period 0, the portfolio mix is 80% A and 20% B, with premiums of $500 and $1,000. In period 1, the mix is 50% A and 50% B, with premiums of $550 and $1,100. Decompose the average premium change using the period 0 mix for the rate effect, and the period 1 rates for the mix effect.

Method. Calculate each period’s weighted average, then apply the first exact decomposition. The common mistake avoided here is calling the whole average increase a rate-level change.

Period 0 average premium is $0.80($500)+0.20($1{,}000)=$400+$200=$600$. Period 1 average premium is $0.50($550)+0.50($1{,}100)=$275+$550=$825$. The observed change is $$825-$600=$225$ per policy.

At the period 0 mix, the rate effect is $0.80($550-$500)+0.20($1{,}100-$1{,}000)=0.80($50)+0.20($100)=$40+$20=$60$.

At the period 1 rates, the mix effect is $(0.50-0.80)($550)+(0.50-0.20)($1{,}100)=(-0.30)($550)+(0.30)($1{,}100)=-$165+$330=$165$.

The effects sum to $$60+$165=$225$, matching the observed change. Thus $60 of the increase is the fixed-mix rate effect under this ordering, while $165 is attributed to the shift in mix at period 1 rates. The result checks against the calculated period averages.

Why the Order of Attribution Matters

The two sequential decompositions are both correct, but they assign the interaction differently. This matters when rates and mix change at the same time. A report should state which ordering it uses; otherwise, two analysts may report different rate and mix contributions while agreeing on the total change. If a neutral allocation is preferred, the symmetric decomposition averages the two sequential allocations.

Worked Example: A Loss Average Rises While Segment Loss Rates Fall

A hypothetical portfolio has two segments, Standard and Coastal. In period 0, 75% of exposure is Standard and 25% Coastal; loss per exposure is $200 and $800, respectively. In period 1, the mix is 40% Standard and 60% Coastal; the segment loss amounts are $190 and $760 per exposure. Period 0 has 1,000 exposures and period 1 has 1,200. Find the change in average loss per exposure and attribute it to rate and mix using the period 0 mix for the rate effect.

Method. Decompose the average loss per exposure first, then calculate the total losses to show the separate exposure-volume effect. The trap is to interpret a higher aggregate loss total as evidence that segment loss amounts increased.

The period 0 average loss is $0.75($200)+0.25($800)=$150+$200=$350$ per exposure. The period 1 average is $0.40($190)+0.60($760)=$76+$456=$532$ per exposure. The average therefore rises by $$532-$350=$182$.

The rate effect at the period 0 mix is $0.75($190-$200)+0.25($760-$800)=0.75(-$10)+0.25(-$40)=-$7.50-$10.00=-$17.50$ per exposure.

The mix effect at the period 1 rates is $(0.40-0.75)($190)+(0.60-0.25)($760)=(-0.35)($190)+(0.35)($760)=-$66.50+$266.00=$199.50$ per exposure. Together, $-$17.50+$199.50=$182.00$, which checks against the average change. Each segment’s loss amount fell by 5%, but the shift toward Coastal exposure more than offset those reductions in the portfolio average.

Total losses were $1{,}000($350)=$350{,}000$ in period 0 and $1{,}200($532)=$638{,}400$ in period 1, a change of $288,400. To separate volume using the later average, the change is $1{,}000($182)+ (1{,}200-1{,}000)($532)=$182{,}000+$106{,}400=$288{,}400$. The first component is the average loss change applied to the earlier exposure count; the second is the additional 200 exposures valued at the later average. This confirms that aggregate dollars combine average and exposure-volume changes.

Symmetric Attribution and a Three-Segment Check

A symmetric decomposition is useful when stakeholders want a rate and mix allocation without selecting which change is applied first. It is exact, but it is still an attribution convention: the two effects are not independent causal estimates. The segment-level values and weights should be shown so reviewers can understand what drives each contribution.

Worked Example: Using a Symmetric Decomposition

A hypothetical premium portfolio has three segments. Segment premiums move from $300, $600, and $1,200 to $330, $540, and $1,320. The mix moves from 50%, 30%, and 20% to 40%, 40%, and 20%. Calculate the symmetric rate and mix effects on average premium.

Method. Apply the average of the two period weights to each premium change for the rate effect, and the average of the two period premiums to each weight change for the mix effect. This avoids the common mistake of treating one arbitrary calculation order as the only possible allocation.

The period 0 average is $0.50($300)+0.30($600)+0.20($1{,}200)=$150+$180+$240=$570$. The period 1 average is $0.40($330)+0.40($540)+0.20($1{,}320)=$132+$216+$264=$612$. The total change is $$42$.

The average weights are 45%, 35%, and 20%. Segment premium changes are $+$30$, $-$60$, and $+$120$. The symmetric rate effect is $0.45($30)+0.35(-$60)+0.20($120)=$13.50-$21.00+$24.00=$16.50$.

The average premiums are $315, $570, and $1,260. The weight changes are $-10\%$, $+10\%$, and 0%. The symmetric mix effect is $(-0.10)($315)+(0.10)($570)+(0)($1{,}260)=-$31.50+$57.00+$0=$25.50$. The effects sum to $$16.50+$25.50=$42.00$, matching the observed average change.

Practical Interpretation and Common Traps

The decomposition is only as meaningful as its basis. Use a consistent exposure unit, segment definition, time period, and amount basis. If a segment has no exposure in one period, its observed segment average may be undefined; a decomposition then requires a defensible rate or loss estimate for that missing cell, or a revised comparison set. Do not silently treat an undefined amount as zero.

Do not confuse average and total. Total premium or loss changes when the number of exposures changes, even if the per-exposure average does not. Report volume separately or standardize totals to a common exposure count.
Do not call the mix effect a rate effect. A changing share of high-premium or high-loss segments can move the aggregate even when every segment amount is unchanged.
State the decomposition order. A sequential rate effect depends on whether it is valued at earlier or later mix. Use the symmetric form if the chosen ordering is not intended to privilege either period.
Check the arithmetic identity. Rate effect plus mix effect must equal the observed change in the average. A mismatch usually indicates inconsistent weights, units, or signs.
Do not infer causation from attribution. The decomposition describes how observed averages differ under a stated calculation. It does not establish why segment rates or the portfolio mix changed.

Key takeaway: Compare portfolio averages on a fixed mix to isolate segment-level changes. Then show the mix contribution and, for aggregate dollars, the exposure-volume contribution. State the basis and attribution order, and verify that the components sum to the observed change.

Check Your Understanding

For each problem, calculate the requested comparison and check that its components reconcile.

A two-segment portfolio has earlier mix 60% A and 40% B, with premiums of $400 and $800. Later premiums are $420 and $840. Find the rate effect at the earlier mix. (Answer: $28 per policy.)
Segment loss amounts remain $250 for A and $750 for B. The share of A changes from 80% to 50%. Find the change in average loss per exposure and identify its source. (Answer: +$150 per exposure, entirely due to mix.)
Weights change from 70%/30% to 40%/60%. Premiums change from $100/$500 to $120/$450. Calculate the symmetric rate effect, symmetric mix effect, and total average change. (Answer: rate effect $-$11.50$, mix effect $+$109.50$, total change $+$98.00$ per policy.)
Average premium rises from $300 to $330 per exposure while exposure count rises from 1,000 to 1,100. Decompose the total premium change by applying the average change to the earlier exposure count and valuing added exposures at the later average. (Answer: total increase $63,000; average-change component $30,000 and volume component $33,000.)
For segments A and B, earlier mix is 75%/25% and later segment loss amounts are $110/$360. Earlier amounts were $100/$400. Using the earlier mix, calculate the rate effect on average loss. (Answer: $-$2.50$ per exposure.)

What You'll Learn