Sensitivity Of Expected Ultimate To Loss Ratio Selection

Why Loss Ratio Selection Sensitivity Matters

A premium-based expected ultimate loss estimate depends directly on the selected expected loss ratio. When the premium measure is held fixed, a higher selection produces a higher expected ultimate; a lower selection produces a lower one. The size of that change is useful to quantify: it shows whether a modest selection difference has a small dollar effect or a material effect on the ultimate and the unpaid estimate.

This sensitivity is not, by itself, evidence that one selection is better than another. It isolates the arithmetic effect of changing the loss ratio while holding other inputs constant. The premium basis, the loss and adjustment expense (LAE) scope, and the period being estimated must remain consistent for the comparison to be meaningful.

Definition: Let $P$ be the premium amount on the basis appropriate for the period and loss measure, and let $r$ be the selected expected loss ratio expressed as a decimal. The premium-based expected ultimate loss is $U=Pr$. For two selections $r_1$ and $r_2$ applied to the same $P$, the change in expected ultimate is $\Delta U=P(r_2-r_1)$. If $C$ is paid loss and LAE to date, expected unpaid loss is $U-C$. If $R$ is reported incurred loss and LAE, the corresponding expected IBNR estimate is $U-R$.

In these definitions, $P$, $C$, and $R$ must be measured in compatible dollars and must refer to the same period and the same loss and LAE scope. For example, a selection intended to include LAE should not be applied to premium to estimate loss-only ultimate and then compared with a loss-and-LAE amount. The applicable premium may be earned premium or another premium measure appropriate to the analysis; the key is that its basis matches the selected ratio.

How the Sensitivity Works

The calculation is linear in the selected ratio. If premium is fixed, increasing the ratio by one percentage point means increasing $r$ by $0.01$, not multiplying the prior ultimate by one percent. The dollar increase is therefore $0.01P$. For example, if the applicable premium is $10 million, one percentage point changes the expected ultimate by $100,000.

U(r)=Pr,\qquad \frac{\partial U}{\partial r}=P,\qquad U(r+\Delta r)-U(r)=P\Delta r.

The derivative states the dollar sensitivity per unit change in the ratio. Since a one-percentage-point change is $0.01$ units, the corresponding dollar change is $0.01P$. This is an exact difference, not an approximation, because $U=Pr$ is linear in $r$.

Derivation and Interpretation

The expected loss ratio is the expected loss amount per dollar of the specified premium. Applying a ratio $r$ to premium $P$ therefore gives the expected loss amount $Pr$. For a second selection $r+\Delta r$, the estimate is $P(r+\Delta r)$. Subtracting the first estimate gives:

P(r+\Delta r)-Pr =Pr+P\Delta r-Pr =P\Delta r.

If paid loss and LAE $C$ are unchanged between the two calculations, subtracting $C$ from each ultimate does not change the dollar difference:

\bigl[P(r+\Delta r)-C\bigr]-\bigl[Pr-C\bigr] =P\Delta r.

Thus, with fixed premium and fixed paid-to-date amounts, the change in expected unpaid loss equals the change in expected ultimate loss. The same cancellation applies to expected IBNR calculated as ultimate less reported incurred: if $R$ is fixed, changing $r$ changes $U-R$ by $P\Delta r$. These statements describe sensitivity to the selection only; they do not establish that paid or reported losses are complete or that the selected ratio is appropriate.

Key result: Holding the premium and loss data fixed, each one-percentage-point change in the selected expected loss ratio changes expected ultimate and expected unpaid loss by $0.01P$. The percentage change in ultimate is not generally one percentage point: relative to a starting selection $r$, it is $\Delta r/r$.

Worked Scenarios

Worked Example: Dollar Impact of a Revised Selection

A hypothetical insurer is estimating ultimate loss and LAE for one accident period. The applicable earned premium is $8,000,000. Paid loss and LAE are $3,100,000. Compare expected ultimate and expected unpaid loss at selected expected loss ratios of 62% and 65%. Quantify the sensitivity and identify the common mistake avoided.

Method. Apply each ratio to the same premium, then subtract the same paid amount from each ultimate. The ratios in decimal form are $r_1=0.62$ and $r_2=0.65$.

At 62%, expected ultimate is $U_1=\$8{,}000{,}000(0.62)=\$4{,}960{,}000$. Expected unpaid loss is $\$4{,}960{,}000-\$3{,}100{,}000=\$1{,}860{,}000$.

At 65%, expected ultimate is $U_2=\$8{,}000{,}000(0.65)=\$5{,}200{,}000$. Expected unpaid loss is $\$5{,}200{,}000-\$3{,}100{,}000=\$2{,}100{,}000$.

The selection change is $0.65-0.62=0.03$, or 3 percentage points. Its dollar effect is $\Delta U=\$8{,}000{,}000(0.03)=\$240{,}000$. The expected unpaid estimate also increases by $\$2{,}100{,}000-\$1{,}860{,}000=\$240{,}000$. The calculation checks because both differences agree with $P\Delta r$.

Common mistake avoided: Increasing the 62% ultimate by 3% would give $\$4{,}960{,}000(1.03)=\$5{,}108{,}800$, which is not the 65% estimate. The ratio increased by 3 percentage points, so the correct estimate is premium times 0.65, or $5,200,000.

Worked Example: Low, Central, and High Selections

For a hypothetical liability portfolio, the premium basis is $12,500,000 and paid loss and LAE are $5,400,000. An analyst wants to show expected unpaid losses at 58%, 62%, and 66%. Calculate all three estimates, the dollar change per percentage point, and the relative increase in ultimate from 58% to 66%. Name the method and the trap it avoids.

Method. This is a fixed-premium sensitivity scenario. Calculate $U=Pr$ for each selection, then calculate expected unpaid as $U-C$.

At 58%, $U=\$12{,}500{,}000(0.58)=\$7{,}250{,}000$, so expected unpaid is $\$7{,}250{,}000-\$5{,}400{,}000=\$1{,}850{,}000$.

At 62%, $U=\$12{,}500{,}000(0.62)=\$7{,}750{,}000$, so expected unpaid is $\$7{,}750{,}000-\$5{,}400{,}000=\$2{,}350{,}000$.

At 66%, $U=\$12{,}500{,}000(0.66)=\$8{,}250{,}000$, so expected unpaid is $\$8{,}250{,}000-\$5{,}400{,}000=\$2{,}850{,}000$.

One percentage point changes the estimate by $0.01(\$12{,}500{,}000)=\$125{,}000$. The 58%-to-66% change is 8 percentage points, so its dollar effect is $8(\$125{,}000)=\$1{,}000{,}000$, matching $\$8{,}250{,}000-\$7{,}250{,}000$. The relative increase in ultimate is $\$1{,}000{,}000/\$7{,}250{,}000=0.137931\ldots$, or 13.79% to two decimals.

Common mistake avoided: The 8-percentage-point change is not an 8% relative increase. The relative increase is measured against the starting ultimate, and is about 13.79% here. The dollar sensitivity check confirms the scenario calculations.

Worked Example: Selection Sensitivity and Negative Expected IBNR

A hypothetical specialty portfolio has applicable premium of $10,500,000, paid loss and LAE of $4,600,000, and reported incurred loss and LAE of $7,900,000. Compare expected ultimate, expected unpaid loss, and expected IBNR at 70% and 74%. Explain how to interpret any negative result. State the basis assumption and the mistake avoided.

Method and basis assumption. Assume both selections and all three dollar amounts refer to the same period and include the same loss and LAE scope. Calculate ultimate as premium times the selected ratio; calculate expected unpaid as ultimate less paid, and expected IBNR as ultimate less reported incurred.

At 70%, $U=\$10{,}500{,}000(0.70)=\$7{,}350{,}000$. Expected unpaid is $\$7{,}350{,}000-\$4{,}600{,}000=\$2{,}750{,}000$. Expected IBNR is $\$7{,}350{,}000-\$7{,}900{,}000=-\$550{,}000$.

At 74%, $U=\$10{,}500{,}000(0.74)=\$7{,}770{,}000$. Expected unpaid is $\$7{,}770{,}000-\$4{,}600{,}000=\$3{,}170{,}000$. Expected IBNR is $\$7{,}770{,}000-\$7{,}900{,}000=-\$130{,}000$.

The four-percentage-point change implies $\$10{,}500{,}000(0.04)=\$420{,}000$ more ultimate. Expected unpaid also rises by $\$420{,}000$, and expected IBNR rises from $-\$550{,}000$ to $-\$130{,}000$, also a $420,000 change. Each calculation reconciles.

Interpretation and common mistake avoided: Both selections produce negative expected IBNR because each selected ultimate is below reported incurred. Do not silently replace negative IBNR with zero: that would change the calculation and conceal the implication of the chosen ultimate. A negative result is a diagnostic signal to review the selection and the alignment of the inputs; it is not, by itself, proof of a basis error.

Using Sensitivity Results Carefully

A sensitivity calculation is useful when evaluating how much judgment in the expected loss ratio matters to the dollar estimate. It can show the range of expected ultimates implied by plausible selections and whether the resulting expected unpaid amounts move materially. It does not select the ratio, assign probabilities to alternative selections, or establish that the calculated range captures all uncertainty.

The fixed-premium result also clarifies what is not being tested. If both premium and the expected loss ratio change, the difference between the two estimates is not simply the old premium times the ratio change. For a move from $(P_1,r_1)$ to $(P_2,r_2)$, the total difference is $P_2r_2-P_1r_1$. A useful exact decomposition is:

$$ P_2r_2-P_1r_1 =P_1(r_2-r_1)+r_2(P_2-P_1). $$

This decomposition first changes the ratio at the original premium, then changes premium using the new ratio. Other decompositions are possible, but the allocation between the two effects depends on the order chosen. For a pure loss ratio sensitivity, hold premium fixed so that this attribution issue does not arise.

Before interpreting a result, verify that the compared selections use the same premium measure and the same loss and LAE scope. Confirm that paid or reported amounts are held constant when describing the effect on unpaid loss or IBNR. Finally, distinguish percentage points from relative percent changes: moving from 60% to 63% is a 3-percentage-point increase, while the ratio itself increases by $3/60=5\%$ relative to its starting value.

Key takeaway: With premium fixed, the exact dollar sensitivity of a premium-based expected ultimate to a selected expected loss ratio is $P\Delta r$. The expected unpaid estimate changes by the same amount when paid-to-date data are fixed. Meaningful interpretation still depends on a consistent period, premium basis, and loss and LAE scope.

Check Your Understanding

For each original scenario, calculate the requested quantities before checking the answer.

A hypothetical account has applicable premium of $6,400,000 and paid loss and LAE of $2,700,000. What are expected ultimate and expected unpaid at 55% and 57.5%, and what is the dollar sensitivity? (Answer: At 55%, ultimate is $3,520,000 and unpaid is $820,000; at 57.5%, ultimate is $3,680,000 and unpaid is $980,000; sensitivity is an increase of $160,000.)
A hypothetical program has premium of $9,000,000. By how much does its premium-based expected ultimate change when the selected ratio moves from 63% to 60%, with premium fixed? (Answer: It decreases by $270,000.)
For a hypothetical portfolio, premium is $15,000,000, paid loss and LAE are $8,200,000, and the selected ratio rises from 52% to 54%. Calculate expected unpaid at both selections and check the change using the sensitivity formula. (Answer: Expected unpaid is -$400,000 at 52% and -$100,000 at 54%; it increases by $300,000, consistent with $0.02(\$15{,}000{,}000)=\$300{,}000$.)
A hypothetical book has premium of $7,500,000 and reported incurred loss and LAE of $4,900,000. Calculate expected IBNR at 64% and 68%, and explain the change. (Answer: Expected IBNR is -$100,000 at 64% and $200,000 at 68%; it increases by $300,000, the premium times the 4-percentage-point change.)
A hypothetical estimate changes from premium of $11,000,000 at a 59% ratio to premium of $10,000,000 at a 62% ratio. Calculate both ultimates and the total change; explain why this is not a fixed-premium ratio sensitivity. (Answer: The first ultimate is $6,490,000 and the second is $6,200,000, a decrease of $290,000; both premium and ratio changed, so the total change cannot be attributed to the ratio alone.)

What You'll Learn