Adjusting Burning Costs For Portfolio Changes

Why Portfolio Changes Require Recasting

A historical burning cost is informative only to the extent that the historical portfolio resembles the portfolio to be priced. A change in the number of exposure units affects the total expected loss, while a change in limits, deductibles, business mix, or underwriting composition can change the expected loss per unit as well. Treating all of these as a simple exposure-volume adjustment can produce a misleading indication.

The starting point is historical layer loss after any required development and trend adjustments. The next step is to express those losses on the prospective coverage and portfolio basis. The earlier material on developing and trending historical layer losses addresses the time adjustment; this tutorial focuses on changes in the portfolio itself.

Definition: A portfolio-adjusted burning-cost indication is an estimate of prospective layer loss per exposure unit, or in total, after historical losses have been adjusted to reflect prospective coverage terms, exposure volume, business mix, and underwriting composition. A change to exposure volume alone scales total losses only when the expected loss per unit is otherwise unchanged.

A useful sequence is to identify the prospective exposure base, adjust individual loss payments to prospective coverage terms, calculate loss costs for comparable risk groups, and then weight those costs by the prospective portfolio distribution. Apply a trend factor before recalculating a coverage payment when severity inflation changes the loss amount to which the deductible, limit, or attachment applies.

Fix the basis.
State the exposure unit, coverage terms, and portfolio composition to which the indication should apply.

Restate losses.
Develop and trend losses as needed, then calculate what each sufficiently observed loss would pay under prospective terms.

Normalize and reweight.
Calculate costs per exposure unit for comparable groups, then weight them using prospective—not historical—exposure shares.

Scale to prospective volume.
Multiply the adjusted cost per unit by prospective exposure volume, provided the unit and risk basis are consistent.

Volume Is Not the Same as Mix

Let $L_j$ denote adjusted layer losses for group $j$, and $E_j$ its exposure units. The historical group loss cost is $r_j=L_j/E_j$. If the prospective portfolio has exposure $E_j^{*}$ in that same group and the group’s risk characteristics and coverage basis are stable, expected loss is estimated as $r_j E_j^{*}$. The total is the sum across groups.

\widehat{L}^{*}=\sum_j r_j E_j^{*}, \qquad \widehat{r}^{*}=\frac{\widehat{L}^{*}}{\sum_j E_j^{*}} =\sum_j w_j^{*}r_j, \qquad w_j^{*}=\frac{E_j^{*}}{\sum_k E_k^{*}}.

The equation shows why a changed portfolio mix matters: the prospective average loss cost is a weighted average of group costs, with prospective exposure shares as weights. Historical aggregate losses divided by historical aggregate exposure instead use historical shares. That calculation is not automatically a suitable prospective rate when the shares change.

Worked Example: Separating Exposure Growth From Loss Cost

A hypothetical excess portfolio has losses already developed and trended to the prospective cost level. Across two years, adjusted losses are $200,000 and $200,000, with 100 and 200 exposure units, respectively. Assume the exposure definition and coverage terms are unchanged, and the prospective book is expected to have 450 units. Calculate the historical burning cost per unit and the indicated prospective total loss.

Method. Combine adjusted losses and exposure units to obtain the historical cost per unit, then apply that cost to prospective volume. The assumption of unchanged risk composition and terms is essential.

Total adjusted losses are $200{,}000+200{,}000=\$400{,}000$. Total historical exposure is $100+200=300$ units. Therefore,

r=\frac{\$400{,}000}{300}=\$1{,}333.333\ldots\text{ per unit}.

At 450 prospective units, indicated loss is $450\times \$1{,}333.333\ldots=\$600{,}000$. As a check, the prospective exposure is 1.5 times the historical 300 units, so the total is also $1.5\times \$400{,}000=\$600{,}000$. The cost per unit is reported as $1,333.33; the indicated total is $600,000.00. The common mistake is to divide by the prospective 450 units when calculating the historical rate, which would mix different exposure bases.

Recalculating Losses Under Changed Terms

A deductible, attachment, or limit changes the payment generated by a ground-up loss. For a loss amount $X$, a layer attaching at $a$ with limit $u$ has payment

Y^L=\min\big((X-a)^+,u\big), \qquad (z)^+=\max(z,0).

This payment function applies when the layer is per loss, attaches at $a$, and is capped at $u$. If losses are recorded only as amounts paid under old terms, the underlying ground-up amounts may not be recoverable; a new coverage calculation then needs suitable exposure information or an explicit severity assumption. A limit does not reveal the size of losses censored above it.

The reason trend must precede a term change is direct: a deductible or attachment is applied to the loss amount at the relevant cost level. In general, for trend factor $t$, the prospective payment is $Y^L(tX)$, not $tY^L(X)$. The two differ when a loss crosses an attachment or limit.

Worked Example: Repricing Losses at Changed Attachments

A hypothetical portfolio has 500 exposure units and two observed ground-up claims. Both claims are to be trended by 10% to the prospective level. Claim A is subject to a prospective layer attaching at $300,000, with no limit reached by this claim. Claim B is subject to a distinct prospective layer attaching at $700,000, also with no limit reached. The second attachment represents a $300,000 retention plus a $400,000 underlying inuring layer. Calculate the combined adjusted loss cost per exposure unit.

Method. Trend each ground-up claim first, then apply the attachment applicable to that claim’s coverage group. This avoids treating different underlying structures as if they had a common attachment.

Claim A trends to $\$450{,}000\times1.10=\$495{,}000$. Its layer payment is $\$495{,}000-\$300{,}000=\$195{,}000$.

Claim B trends to $\$900{,}000\times1.10=\$990{,}000$. Its target layer begins above the combined $700,000 underlying amount, so its payment is $\$990{,}000-\$300{,}000-\$400{,}000=\$290{,}000$. This claim does not reach a cap under the stated terms.

Combined adjusted losses are $\$195{,}000+\$290{,}000=\$485{,}000$. Dividing by 500 exposure units gives $\$485{,}000/500=\$970$ per unit. The check is that the two layer payments total $485,000, and multiplying the unrounded unit cost by 500 returns $485,000. The common mistake is to use the same attachment for both claims or to trend an old layer payment rather than trend ground-up severity and then apply the prospective terms.

Reweighting Business Mix and Underwriting Composition

Where groups have different loss costs, an overall indication should reflect the prospective distribution of exposures among those groups. This applies both to broad business segments and to rating or underwriting categories when they materially distinguish expected layer loss. The weights must use the exposure measure appropriate to the indicated cost.

Worked Example: Changing the Segment Mix

A hypothetical insurer has two segments with adjusted layer costs of $800 and $2,000 per exposure unit. Historical exposure was 700 units in Segment A and 300 in Segment B. The prospective portfolio is expected to have 400 units in A and 600 in B, for 1,000 total units. Assume costs within each segment are stable and all losses are on the same prospective coverage and trend basis. Find the historical blended cost, prospective blended cost, and prospective total loss indication.

Method. Calculate each blended cost with the relevant exposure shares. Do not carry forward the historical blend when prospective shares differ.

Historical blend: Segment A contributes $700\times\$800=\$560{,}000$, and Segment B contributes $300\times\$2{,}000=\$600{,}000$. Total historical adjusted losses are $1,160,000 over 1,000 units, or $\$1{,}160$ per unit.

Prospective blend: Segment A contributes $400\times\$800=\$320{,}000$, and Segment B contributes $600\times\$2{,}000=\$1{,}200{,}000$. Total prospective loss is $\$320{,}000+\$1{,}200{,}000=\$1{,}520{,}000$. Dividing by 1,000 prospective units gives $\$1{,}520$ per unit.

The prospective cost is $360 per unit higher than the historical blend because the higher-cost segment’s share rose from 30% to 60%. The calculation checks: $0.40(\$800)+0.60(\$2{,}000)=\$1{,}520$. Using the historical $1,160 blend for the new book would understate the indication.

Underwriting composition can also change inside a broad segment. For example, a book may contain a different share of high-limit locations, higher-hazard operations, or risks meeting stricter selection criteria. If those categories have different expected layer costs, calculate costs on a sufficiently comparable basis and reweight using prospective composition. If an underwriting change alters the risk characteristics within the available categories, historical category costs may no longer represent prospective costs; the analyst needs a defensible adjustment or a more appropriate basis rather than an unsupported assumption of stability.

Practical Checks and Limits of the Adjustment

A portfolio adjustment is not a substitute for adequate data. Recalculating claims under changed terms requires loss information detailed enough to apply the prospective payment function. Losses censored by an old limit do not identify the full severity above that limit, and losses net of an old deductible may not identify ground-up amounts below it. Any severity model or exposure curve used to fill that gap must be appropriate to the risks and must be stated; do not treat unobserved amounts as known claims.

Before relying on a recast burning cost, check that development, trend, coverage, and exposure adjustments are not counted twice. Confirm that exposure units mean the same thing in the historical and prospective denominators. Separate changes in total volume from changes in the shares and characteristics of risks. Finally, compare the adjusted total with the sum of adjusted group losses and the adjusted cost per unit multiplied by prospective exposure.

Do not scale totals alone when class mix or coverage terms changed.
Do not apply limits or deductibles in the wrong order. Trend ground-up losses before applying prospective terms when the trend changes severity.
Do not infer uncensored severity from capped payments. State what additional data or assumption supports the recast.
Do not apply a group rate indiscriminately if underwriting composition within that group has materially changed.

Key takeaway: A prospective burning-cost indication is built by bringing losses onto prospective coverage terms, calculating comparable costs per exposure unit, and applying prospective—not historical—volume and mix. Scale total losses by exposure only when the underlying expected cost per unit is stable.

Check Your Understanding

For each problem, assume the stated losses and rates are on the required prospective trend basis unless a trend is specified.

A homogeneous hypothetical book has $540,000 of adjusted losses over 360 exposure units. If prospective exposure is 450 units and risk and terms are unchanged, find the indicated cost per unit and total loss. (Answer: $1,500 per unit; $675,000 total.)
A claim of $260,000 is trended by 5%. Under a prospective $250,000 attachment with no limit reached, find the layer payment. (Answer: $23,000.)
Two segments have costs of $600 and $1,500 per unit. Prospective shares are 75% and 25%, with 800 total units. Find the blended cost and total indicated loss. (Answer: $825 per unit; $660,000 total.)
A layer attaches at $400,000 and has a $300,000 limit. A ground-up claim is $850,000 before a 10% trend. Find its prospective layer payment. (Answer: $300,000; the trended loss is $935,000 and the uncapped layer amount is $535,000.)
Historical adjusted losses are $720,000 over 600 units. A prospective book has 900 units, but the higher-cost segment’s share rises from 20% to 50%. Explain why multiplying $720,000 by 900/600 is not necessarily a valid indication. (Answer: It assumes the historical average cost per unit remains valid; the changed segment share requires segment-level costs reweighted to prospective shares.)

What You'll Learn