Portfolio Mix Shifts and Aggregate Reserve Comparisons

Why Portfolio Mix Matters

An aggregate reserve can change even when every segment’s reserve estimate per unit stays the same. The reason is that the portfolio may contain different proportions of segments with different reserve estimates. For example, shifting exposures toward a line with a higher selected unpaid estimate per exposure increases the aggregate reserve, all else equal.

This distinction matters when interpreting reserve comparisons. A change in the total can reflect a different portfolio composition rather than a change in segment-level assumptions. The calculations below isolate that composition effect; they do not establish why the mix changed or whether the underlying estimates are adequate. The treatment is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Definition: Let $n_{it}$ be the number of units in segment $i$ at time $t$, and let $r_i$ be the selected unpaid reserve per unit for that segment. A unit might be an exposure, an open claim, or another consistently defined quantity. If the segments are mutually exclusive and collectively exhaustive, the aggregate reserve is $R_t=\sum_i n_{it}r_i$. The segment estimate $r_i$, the unit definition, and the reserve basis must be consistent across the comparison.

If the total number of units is $N_t=\sum_i n_{it}$, define the segment weight as $w_{it}=n_{it}/N_t$. The weights are proportions, so $\sum_i w_{it}=1$. The aggregate reserve can then be written as

R_t=N_t\sum_i w_{it}r_i

The weighted average $\sum_i w_{it}r_i$ is the reserve per unit for the portfolio as a whole. It is not generally the simple average of the segment estimates: a segment’s contribution depends on both its estimate per unit and its share of the portfolio.

Deriving the Mix Effect

First hold total units fixed at $N$, and suppose the segment estimates $r_i$ do not change. Comparing an initial mix $w_{i0}$ with a later mix $w_{i1}$, the reserve difference is

\Delta R =N\sum_i w_{i1}r_i-N\sum_i w_{i0}r_i =N\sum_i (w_{i1}-w_{i0})r_i

This follows by subtracting the initial weighted sum from the later weighted sum and factoring out $N$. Because both sets of weights sum to one, $\sum_i(w_{i1}-w_{i0})=0$. Thus adding the same constant to every segment estimate would not change the calculated mix effect: the result depends on how the shifted weights relate to differences among the segment estimates.

For two segments, if a proportion $q$ moves from segment L to segment H, the weight change is $-q$ for L and $+q$ for H. Substitution gives

\Delta R=N[-q r_L+q r_H]=Nq(r_H-r_L)

So moving units toward the higher-reserve segment increases the aggregate by the number of shifted units multiplied by the difference in reserve per unit. Moving them toward the lower-reserve segment decreases it. If all segments have the same $r_i$, a change in weights alone has no effect.

Condition: This is a pure mix comparison only when total units and all segment-level reserve estimates are held fixed. If the total number of exposures or claims also changes, or if any $r_i$ changes, the observed aggregate difference includes more than a mix effect.

Worked Comparisons

Worked Example: Two Lines With Different Reserve Estimates

A hypothetical insurer has 500 exposures in each of two periods. Line X has an unpaid estimate of $90 per exposure, and Line Y has an unpaid estimate of $210 per exposure. X represents 80% of exposures initially and 50% later. Calculate both aggregate reserves and the change attributable to mix.

Method. Use the weighted-reserve formula, with exposure counts as the units. The total is fixed at 500 and both per-exposure estimates are fixed.

Initial period. X has $500(0.80)=400$ exposures, and Y has $500(0.20)=100$. Their reserve contributions are $400\times \$90=\$36{,}000$ and $100\times \$210=\$21{,}000$. Therefore, the initial aggregate is $\$36{,}000+\$21{,}000=\$57{,}000$.

Later period. X has $500(0.50)=250$ exposures, and Y has $500(0.50)=250$. Their contributions are $250\times \$90=\$22{,}500$ and $250\times \$210=\$52{,}500$. The later aggregate is $\$22{,}500+\$52{,}500=\$75{,}000$.

The mix-related increase is $\$75{,}000-\$57{,}000=\$18{,}000$. As a check, 150 exposures shifted from X to Y, and the reserve difference per shifted exposure is $\$210-\$90=\$120$. Thus $150\times \$120=\$18{,}000$, matching the aggregate calculation.

Common mistake avoided: The initial reserve is not the total exposure count multiplied by an unweighted average of $90 and $210. The 80% and 20% weights matter. Ignoring them would give an incorrect initial reserve.

Worked Example: A Shift in Open-Claim Types

A hypothetical book has 1,500 open claims at each valuation. Type A has an unchanged selected reserve of $1,200 per claim, and Type B has an unchanged selected reserve of $4,200 per claim. Type A’s share falls from 60% to 40%. Calculate the aggregate at each valuation and the mix effect.

Method. Apply the per-claim estimates to the counts of each claim type. Because the total claim count is unchanged, the difference isolates the effect of mix under the stated assumptions.

Initial valuation. There are $1{,}500(0.60)=900$ Type A claims and $1{,}500(0.40)=600$ Type B claims. The reserve is $900\times \$1{,}200=\$1{,}080{,}000$ for A and $600\times \$4{,}200=\$2{,}520{,}000$ for B. The aggregate is $\$1{,}080{,}000+\$2{,}520{,}000=\$3{,}600{,}000$.

Later valuation. There are $1{,}500(0.40)=600$ Type A claims and $1{,}500(0.60)=900$ Type B claims. The contributions are $600\times \$1{,}200=\$720{,}000$ and $900\times \$4{,}200=\$3{,}780{,}000$. The later aggregate is $\$720{,}000+\$3{,}780{,}000=\$4{,}500{,}000$.

The increase is $\$4{,}500{,}000-\$3{,}600{,}000=\$900{,}000$. Equivalently, 300 claims shifted from A to B, so the increase is $300(\$4{,}200-\$1{,}200)=300(\$3{,}000)=\$900{,}000$. Both calculations agree.

Common mistake avoided: Keep each reserve estimate attached to its own claim type. Using Type A’s estimate for the later Type B claims, or reversing the type counts, changes the answer and misstates the mix effect.

Worked Example: Three Territories and a Weighted Check

A hypothetical portfolio has 800 exposures in both periods. The selected reserve per exposure is $45 in Territory 1, $120 in Territory 2, and $300 in Territory 3. The initial shares are 50%, 30%, and 20%; the later shares are 35%, 40%, and 25%. Calculate the aggregate reserves and explain the change.

Initial period. The exposure counts are $800(0.50)=400$, $800(0.30)=240$, and $800(0.20)=160$. The respective reserve contributions are $400\times \$45=\$18{,}000$, $240\times \$120=\$28{,}800$, and $160\times \$300=\$48{,}000$. The initial aggregate is $\$18{,}000+\$28{,}800+\$48{,}000=\$94{,}800$.

Later period. The counts are $800(0.35)=280$, $800(0.40)=320$, and $800(0.25)=200$. The contributions are $280\times \$45=\$12{,}600$, $320\times \$120=\$38{,}400$, and $200\times \$300=\$60{,}000$. The later aggregate is $\$12{,}600+\$38{,}400+\$60{,}000=\$111{,}000$.

The increase is $\$111{,}000-\$94{,}800=\$16{,}200$. A contribution check gives $(280-400)\times \$45=-\$5{,}400$, $(320-240)\times \$120=\$9{,}600$, and $(200-160)\times \$300=\$12{,}000$. Their sum is $-\$5{,}400+\$9{,}600+\$12{,}000=\$16{,}200$.

A second check uses portfolio averages: $\$94{,}800/800=\$118.50$ initially and $\$111{,}000/800=\$138.75$ later. The increase per exposure is $\$20.25$, and $800\times \$20.25=\$16{,}200$. This confirms both the aggregate calculation and its units.

Common mistake avoided: The average of the three territory estimates is not the portfolio estimate unless the territory shares are equal. Use each period’s actual weights.

When Total Volume Also Changes

A real comparison may involve both a different mix and a different total number of units. Let $A_t=\sum_i w_{it}r_i$ be the average reserve per unit at time $t$, with segment estimates held fixed. Then $R_t=N_tA_t$, and the following exact decomposition uses the initial mix for the volume component:

$$ R_1-R_0=(N_1-N_0)A_0+N_1(A_1-A_0) $$

The first term changes total units while holding the initial average reserve per unit fixed. The second changes the mix at the later total volume. This is an exact identity: expanding the right-hand side gives $N_1A_0-N_0A_0+N_1A_1-N_1A_0=N_1A_1-N_0A_0$. One could instead apply the mix change at the initial volume and then the volume change at the later mix. That alternative is also exact, but it allocates the volume-mix interaction differently. State the convention used when reporting such a decomposition.

Interpretation and Common Traps

A mix effect is an arithmetic explanation of an aggregate change under specified assumptions; it is not, by itself, evidence of reserve strengthening, a release, or a change in claim adequacy. Before attributing a reported difference to mix, check that the units are comparable, the segment definitions cover the portfolio without overlap, and each segment estimate is on the same valuation and presentation basis in both periods. For example, do not combine a gross estimate for one segment with a net estimate for another.

Confusing counts with shares. Shares must sum to one within each period, and counts are obtained by multiplying each share by that period’s total.
Taking an unweighted average. The aggregate per-unit estimate is weighted by segment counts, not by the number of segment labels.
Changing multiple things but calling it pure mix. If total units or segment estimates also change, identify those changes separately or state the decomposition convention.
Ignoring direction. A shift toward a segment with a higher reserve per unit raises the aggregate, all else equal; a shift toward a lower one reduces it.
Over-interpreting a segment average. If a segment’s own composition or reserve estimate changes, treating its $r_i$ as fixed may no longer be appropriate.

Key takeaway: With fixed total units and fixed segment estimates, the aggregate reserve change is the total multiplied by the change in the weighted average reserve per unit. Calculate with period-specific weights, verify the segment contributions, and describe the result as a mix effect only when the other inputs are held constant.

Check Your Understanding

For each problem, assume the stated segment estimates are unchanged and use the same reserve basis in both periods.

A hypothetical portfolio has 600 exposures. Segment A has a reserve estimate of $70 per exposure, and Segment B has $190. A’s share changes from 75% to 55%. Find both aggregate reserves and the mix effect. (Answer: Initial $60,000; later $74,400; increase $14,400.)
A book has 900 open claims. Type 1 has a reserve estimate of $800 per claim and Type 2 has $2,800. Type 1’s share changes from two-thirds to one-half. Find the aggregate at each valuation and the change. (Answer: Initial $1,320,000; later $1,620,000; increase $300,000.)
A portfolio has 1,000 exposures and three segments with reserve estimates of $30, $90, and $210 per exposure. Shares change from 40%, 40%, and 20% to 30%, 50%, and 20%. Calculate the aggregate change and identify the segment contribution that is unchanged. (Answer: Initial $90,000; later $96,000; increase $6,000; the third segment contributes $42,000 in both periods.)
Two segments have fixed estimates of $50 and $150 per exposure. The initial portfolio has 1,000 exposures with shares 70% and 30%. Later it has 1,200 exposures with shares 50% and 50%. Calculate the actual aggregate change, then decompose it by first changing volume at the initial mix and then changing mix at the later volume. (Answer: Initial $80,000; later $120,000; increase $40,000, comprising $16,000 volume effect and $24,000 mix effect under the stated order.)
Two segments each have a fixed reserve estimate of $125 per exposure. Total exposure is 2,000 in both periods, but Segment A’s share changes from 20% to 65%. What is the mix effect, and why? (Answer: $0; the weighted average remains $125 per exposure because the segment estimates are equal.)

What You'll Learn