Comparing Expected And Developed Ultimate Losses

Why Compare Two Ultimate Estimates?

An a priori expected ultimate estimate and a developed ultimate estimate are two separate estimates of the same period’s ultimate losses. Comparing them makes the size and direction of their difference explicit. The comparison does not, by itself, tell us which estimate is preferable or why they differ; it records the result of applying two methods to a common target.

An a priori estimate is based on expectations selected independently of the period’s observed loss development, such as an expected loss ratio applied to premium or an expected loss cost applied to exposure. A developed estimate uses observed losses at a valuation date and a selected development method to estimate ultimate losses. The development method is assumed to have already been applied; the focus here is how to compare its ultimate estimate with the a priori estimate.

Definition: Let $U^E$ denote the a priori expected ultimate loss estimate for a specified period, and let $U^D$ denote the developed ultimate loss estimate for that same period. Define the signed difference as $\Delta=U^D-U^E$. A positive $\Delta$ means the developed estimate is higher; a negative $\Delta$ means it is lower. The percentage difference relative to the a priori estimate is $\Delta/U^E$, provided $U^E>0$.

A meaningful comparison requires both estimates to refer to the same period and to use a consistent loss basis and scope—for example, the same coverage, included loss adjustment expense (LAE) treatment, and currency basis. The estimates must also be in the same units. These conditions define what is being compared; they do not make the two estimates interchangeable.

Calculating the Difference

The a priori estimate may be calculated from premium and an expected loss ratio, or from exposure and an expected loss cost. For example, with premium $P$ and expected loss ratio $r^E$, the expected ultimate is $U^E=P r^E$. With exposure $E$ and expected loss cost $c^E$ per exposure unit, it is $U^E=E c^E$. A developed estimate may be calculated by applying a selected development factor to observed losses at the valuation date.

U^E=P r^E \quad \text{or} \quad U^E=E c^E

\Delta=U^D-U^E, \qquad \text{percentage difference}=\frac{U^D-U^E}{U^E}\times 100\%.

The denominator in the percentage calculation matters. Here it is the a priori estimate, so the percentage answers: “By what percentage is the developed estimate above or below the a priori estimate?” If the question instead specifies the developed estimate as the reference, that denominator must be used and the result will generally differ.

Worked Example: Premium-Based Expectation Versus Developed Ultimate

A hypothetical commercial property portfolio has $2,400,000 of premium on the selected basis for one accident period. Its selected expected loss ratio is 62%. At the valuation date, incurred losses for that period are $1,050,000, and the selected incurred development factor to ultimate is $31/21$. Paid losses to date are $820,000. Compare the two ultimate estimates, and state the developed unpaid amount. Use the a priori estimate as the percentage-difference denominator.

Method and expected ultimate. Apply the expected loss ratio to the stated premium:

U^E=\$2{,}400{,}000(0.62)=\$1{,}488{,}000.

Developed ultimate. Apply the incurred development factor to incurred losses:

U^D=\$1{,}050{,}000\left(\frac{31}{21}\right) =\$50{,}000(31) =\$1{,}550{,}000.

Signed and percentage differences. Subtract the a priori estimate from the developed estimate, then divide by the a priori estimate:

\Delta=\$1{,}550{,}000-\$1{,}488{,}000=\$62{,}000, \qquad \frac{\$62{,}000}{\$1{,}488{,}000}\times100\%=4.1667\%.

The developed ultimate is $62,000, or 4.1667%, higher than the a priori expected ultimate. As a separate calculation, developed unpaid losses are developed ultimate less paid losses to date: $1,550,000 − $820,000 = $730,000.

Check and trap avoided. The factor is greater than 1, so the developed ultimate exceeds the incurred amount: $1,550,000 > $1,050,000. Ultimate less incurred is $500,000, but it is not the unpaid amount. Incurred generally includes paid losses plus case reserves; unpaid is ultimate less paid. Here, the given paid amount makes the developed unpaid calculation possible, and the two quantities must not be mislabeled.

Why the Signed Difference Is Useful

The sign preserves direction. If $U^D>U^E$, then $\Delta>0$; if $U^D<U^E$, then $\Delta<0$. Reporting only the absolute size would hide whether the developed estimate is higher or lower. Reporting both dollars and a percentage adds scale: a $50,000 difference has a different relative size against a $500,000 expectation than against a $5,000,000 expectation.

The percentage difference is a descriptive comparison, not a new estimate. It depends on the selected reference amount, and it is undefined if $U^E=0$. If the a priori estimate is very small, the percentage can be large even when the dollar difference is modest; in that situation, report the dollar amounts and signed difference prominently.

Justification: Since $\Delta=U^D-U^E$, adding $\Delta$ to the a priori estimate recovers the developed estimate: $U^E+\Delta=U^E+(U^D-U^E)=U^D$. Dividing this difference by $U^E>0$ expresses the same dollar difference as a fraction of the a priori estimate; multiplying by 100 converts that fraction to a percentage. Thus the sign gives direction, while the denominator sets the reference scale.

Worked Example: Exposure-Based Expectation Versus Developed Ultimate

A hypothetical roadside-assistance program has 18,000 covered vehicle-years for an accident period. The selected a priori expected loss cost is $74 per vehicle-year. At the valuation date, incurred losses are $1,080,000, and the selected incurred development factor to ultimate is 1.25. Compare the estimates in dollars and as a percentage of the a priori estimate. Round the percentage to four decimal places.

Method and expected ultimate. Multiply the number of exposure units by the expected loss cost per unit:

U^E=18{,}000(\$74)=\$1{,}332{,}000.

Developed ultimate. Multiply incurred losses at the valuation date by the selected factor:

U^D=\$1{,}080{,}000(1.25)=\$1{,}350{,}000.

Compare. The developed estimate exceeds the expected estimate by:

\Delta=\$1{,}350{,}000-\$1{,}332{,}000=\$18{,}000, \qquad \frac{\$18{,}000}{\$1{,}332{,}000}\times100\%=1.3514\%.

The developed ultimate is $18,000, or 1.3514%, above the a priori expected ultimate. As a check, the expected amount is based on 18,000 units at $74 each, while the developed amount is 25% above $1,080,000; both calculations produce amounts in dollars for the specified accident period.

Method and trap avoided. This compares ultimate with ultimate. A common error is to compare the expected ultimate with the observed incurred amount instead of first developing the incurred amount. That would compare $1,332,000 with $1,080,000 and report a difference of −$252,000, which is not the requested developed-ultimate comparison.

Comparing Estimates by Segment

When a portfolio has segments, compare estimates within each segment on a consistent basis before interpreting the total. Segment differences can have opposite signs. Adding the estimates first and comparing the totals gives the net portfolio difference, while the segment results show how that net was formed.

For a complete partition into $n$ segments, let $U^E_j$ and $U^D_j$ be the expected and developed ultimate estimates for segment $j$. If each total is the sum of its segment estimates, then:

\Delta_{\text{total}} =\sum_{j=1}^{n}U^D_j-\sum_{j=1}^{n}U^E_j =\sum_{j=1}^{n}(U^D_j-U^E_j) =\sum_{j=1}^{n}\Delta_j.

This identity follows by distributing subtraction across the sums. It is a useful arithmetic check, but it does not mean that a positive difference in one segment should be offset by changing the estimate for another. Preserve both estimates and their differences.

Worked Example: Offsetting Differences Across Two Segments

A hypothetical insurer compares expected and developed ultimate losses for two coverage segments in the same accident period. For Segment A, premium is $1,200,000 and the expected loss ratio is 60%; its developed ultimate is $780,000. For Segment B, premium is $800,000 and the expected loss ratio is also 60%; its developed ultimate is $410,000. Calculate each expected ultimate, each signed difference, and the aggregate difference. Use each segment’s expected ultimate as its percentage denominator.

Expected estimates. Apply the selected loss ratio to each segment’s premium:

U^E_A=\$1{,}200{,}000(0.60)=\$720{,}000, \qquad U^E_B=\$800{,}000(0.60)=\$480{,}000.

Segment comparisons. Subtract expected from developed in each segment:

\Delta_A=\$780{,}000-\$720{,}000=\$60{,}000, \qquad \frac{\$60{,}000}{\$720{,}000}\times100\%=8.3333\%.

\Delta_B=\$410{,}000-\$480{,}000=-\$70{,}000, \qquad \frac{-\$70{,}000}{\$480{,}000}\times100\%=-14.5833\%.

Aggregate comparison. Sum the expected amounts and developed amounts separately:

U^E_{\text{total}}=\$720{,}000+\$480{,}000=\$1{,}200{,}000,

U^D_{\text{total}}=\$780{,}000+\$410{,}000=\$1{,}190{,}000, \qquad \Delta_{\text{total}}=\$1{,}190{,}000-\$1{,}200{,}000=-\$10{,}000.

The aggregate percentage difference is $-\$10{,}000/\$1{,}200{,}000\times100\%=-0.8333\%$. The check is that the segment differences sum to the aggregate difference: $60,000 + (−$70,000) = −$10,000.

Method and trap avoided. The aggregate developed estimate is $10,000 below the aggregate expected estimate, even though Segment A’s developed estimate is higher. Reporting only the aggregate would conceal the segment-level signs; reporting only Segment A would misstate the portfolio result. The segment estimates remain separate, and their differences reconcile to the total.

Common Mistakes in Ultimate Comparisons

Using inconsistent periods or scopes. A premium-based expected estimate for one period cannot be meaningfully compared with a developed estimate for another. Confirm the period, coverage, loss and LAE scope, and units first.
Reversing the subtraction. With $\Delta=U^D-U^E$, a positive result means developed is higher. State the definition before interpreting the sign.
Changing the percentage denominator without saying so. The percentage in this tutorial uses $U^E$. Dividing by $U^D$ answers a different question.
Calling ultimate less incurred “unpaid.” Unpaid loss is ultimate less paid-to-date. Incurred losses generally include paid losses plus case reserves, so ultimate less incurred is not the unpaid amount.
Blending instead of comparing. An average or weighted combination of the two estimates is a different selection. The comparison should preserve $U^E$, $U^D$, and their difference as distinct quantities.

For example, if $U^E=\$900{,}000$ and $U^D=\$960{,}000$, the signed difference is $+\$60{,}000$, and the percentage difference relative to expected is $60{,}000/900{,}000=6.6667\%$. Averaging the estimates to obtain $930,000 would not be a comparison; it would create a third estimate without a stated selection method.

Key takeaway: State both ultimate estimates, define the subtraction direction, and name the percentage denominator. Compare estimates on a common period and loss basis, keep segment results visible when relevant, and calculate unpaid loss from paid-to-date—not from incurred losses.

Check Your Understanding

For each problem, keep the a priori expected ultimate and developed ultimate separate. Use the a priori estimate as the percentage denominator unless the question says otherwise.

A hypothetical equipment insurer has $3,500,000 of premium and a selected expected loss ratio of 58%. Its developed ultimate is $2,100,000. Calculate the expected ultimate, the signed difference, and the percentage difference. (Answer: Expected ultimate $2,030,000; developed is higher by $70,000, or 3.4483%.)
A hypothetical bicycle-sharing program has 12,500 exposure units and an expected loss cost of $96 per unit. Incurred losses are $960,000 and the selected factor to ultimate is 1.30. Compare the estimates in dollars and as a percentage. (Answer: Expected ultimate $1,200,000; developed ultimate $1,248,000; difference +$48,000, or +4.0000%.)
For a hypothetical accident period, premium is $1,600,000, the selected expected loss ratio is 65%, and the developed ultimate is $1,000,000. Calculate the signed and percentage differences and interpret the sign. (Answer: Expected ultimate $1,040,000; difference −$40,000, or −3.8462%; developed is lower.)
A hypothetical portfolio has Segment X expected and developed ultimates of $450,000 and $490,000, respectively, and Segment Y estimates of $350,000 and $320,000. Calculate the aggregate signed difference and percentage, and verify it from the segment differences. (Answer: Segment differences +$40,000 and −$30,000; aggregate difference +$10,000, or +1.2500% of the $800,000 expected total.)
A hypothetical period has expected ultimate losses of $1,260,000, developed ultimate losses of $1,310,000, paid losses of $700,000, and incurred losses of $890,000. Calculate the ultimate-estimate difference and developed unpaid losses. (Answer: Developed is higher than expected by $50,000; developed unpaid losses are $1,310,000 − $700,000 = $610,000.)

What You'll Learn