Deriving Unpaid Liabilities from Ultimate Losses

Introduction

A projected ultimate loss answers how much a cohort of claims is expected to cost in total over its full lifetime. It does not, by itself, answer how much remains unpaid at a particular valuation date. To estimate the unpaid claim liability, the amount already paid must be removed from the estimated ultimate.

This distinction is central when using a chain ladder ultimate projection. The chain ladder supplies an estimated ultimate loss; the liability calculation then compares that estimate with cumulative payments through the valuation date. The subtraction is simple, but it is only meaningful when both amounts refer to the same cohort, valuation basis, and loss measure.

Definition: For a defined cohort of claims, let $U$ be its ultimate loss on a specified basis, and let $C(t)$ be cumulative paid loss through valuation date $t$, measured on that same basis. The estimated unpaid loss liability at $t$ is $$ \hat R(t)=\hat U-C(t), $$ where $\hat U$ is the selected estimate of ultimate loss. The ultimate estimate is a total-cost estimate; the unpaid liability is the estimated portion of that total not yet paid.

Why Ultimate Less Paid Gives Unpaid Loss

The formula follows directly from how cumulative payments divide between amounts already paid and amounts paid later. First consider the underlying amounts, before estimation. Let $C(T)$ be cumulative paid loss when all payments for the cohort have been made. The ultimate loss is $U=C(T)$. At valuation date $t$, the amount paid after that date is $C(T)-C(t)$. Therefore:

\text{future payments after }t =C(T)-C(t) =U-C(t).

At the valuation date, the true ultimate $U$ is not yet known for an immature cohort. The practical estimate substitutes $\hat U$ for $U$, while the cumulative paid amount through the date is observed. This gives $\hat R(t)=\hat U-C(t)$. The identity does not depend on how $\hat U$ was estimated. In this unit, it may be the chain ladder ultimate estimate from earlier material.

The formula is an estimate of future unpaid loss payments, not a second estimate of the cohort's total cost. It is generally stated on an undiscounted loss basis. If a problem specifies a different basis, such as discounted amounts or net rather than gross losses, both terms must be expressed consistently on that basis.

Important distinction: $\hat U$ includes losses already paid and losses estimated to be paid in the future. $\hat R(t)$ includes only the estimated unpaid portion at $t$. Do not report $\hat U$ itself as the unpaid liability unless no payments have yet been made.

Calculating and Aggregating Unpaid Liabilities

The calculation can be made separately for each accident year, policy year, or other defined cohort. The cohort definition must be the same for its ultimate estimate and its cumulative paid amount. Once each cohort's liability is calculated, the estimates can be summed to obtain the total for the set of cohorts, provided their loss bases are consistent.

\hat R_{\mathrm{total}}(t) =\sum_{i=1}^{n}\hat R_i(t) =\sum_{i=1}^{n}\bigl(\hat U_i-C_i(t)\bigr) =\left(\sum_{i=1}^{n}\hat U_i\right) -\left(\sum_{i=1}^{n}C_i(t)\right).

Here $i$ indexes the cohorts. The aggregate identity holds because addition distributes across the cohort-level differences. It is useful to calculate both the cohort estimates and the aggregate: cohort-level results help identify where the liability comes from, while the aggregate provides the total for the selected portfolio.

Worked Example: Unpaid Liability for Several Accident Years

A hypothetical insurer has selected chain ladder ultimate loss estimates for three accident years. At the current valuation date, cumulative paid losses are available on the same undiscounted gross-loss basis. Calculate the estimated unpaid liability for each year and in total.

Accident year	Estimated ultimate loss	Cumulative paid loss
2022	$1,480,000	$1,110,000
2023	$1,260,000	$756,000
2024	$840,000	$210,000

Method. For each accident year, subtract cumulative paid loss from the estimated ultimate loss.

For 2022, the estimated unpaid amount is $ \$1,480,000-\$1,110,000=\$370,000$.

For 2023, it is $ \$1,260,000-\$756,000=\$504,000$.

For 2024, it is $ \$840,000-\$210,000=\$630,000$.

The aggregate estimated ultimate is $ \$1,480,000+\$1,260,000+\$840,000=\$3,580,000$. Aggregate paid loss is $ \$1,110,000+\$756,000+\$210,000=\$2,076,000$. Thus the aggregate liability is $ \$3,580,000-\$2,076,000=\$1,504,000$.

Check. The cohort liabilities sum to $ \$370,000+\$504,000+\$630,000=\$1,504,000$, matching the aggregate calculation. The common mistake avoided is calling the $3,580,000 estimated ultimate the unpaid liability; it includes the $2,076,000 already paid.

Updating the Estimate at a Later Valuation Date

The unpaid estimate can change between valuation dates for two reasons: additional payments are made, and the selected ultimate estimate may be revised. These effects work in opposite directions when considered separately. More paid loss reduces the remaining unpaid amount if the ultimate estimate is unchanged. An increase in the ultimate estimate increases the unpaid amount if paid loss is unchanged.

Let $t_1$ be an earlier valuation date and $t_2$ a later one. The change in estimated liability is:

\hat R(t_2)-\hat R(t_1) =\bigl(\hat U(t_2)-\hat U(t_1)\bigr) -\bigl(C(t_2)-C(t_1)\bigr).

This follows by subtracting the two liability formulas: $\hat R(t_2)=\hat U(t_2)-C(t_2)$ and $\hat R(t_1)=\hat U(t_1)-C(t_1)$. Rearranging gives the displayed change formula. It makes clear why a liability need not decline by exactly the amount paid during a period: the estimated ultimate may also change.

Worked Example: Payments Emerge While the Ultimate Is Revised

For a hypothetical accident-year cohort, the estimated ultimate at the first valuation date is $920,000 and cumulative paid loss is $575,000. By the next valuation date, another $105,000 has been paid, and the selected ultimate estimate has increased to $1,020,000. Calculate the liability at each date and reconcile the change.

At the first date. The estimated unpaid liability is $ \$920,000-\$575,000=\$345,000$.

At the second date. Cumulative paid loss is $ \$575,000+\$105,000=\$680,000$. The updated liability is $ \$1,020,000-\$680,000=\$340,000$.

The liability has decreased by $ \$340,000-\$345,000=-\$5,000$. Reconcile this using the change formula: the ultimate estimate increased by $ \$1,020,000-\$920,000=\$100,000$, while paid loss increased by $105,000. Therefore the liability change is $ \$100,000-\$105,000=-\$5,000$.

Check. The updated ultimate is exactly the updated paid amount plus the updated unpaid estimate: $ \$680,000+\$340,000=\$1,020,000$. The common mistake avoided is reducing the old liability by the $105,000 payment while ignoring the $100,000 increase in the ultimate estimate.

Use One Consistent Loss Basis

The subtraction is valid only if the ultimate and paid amounts measure the same thing. For example, gross ultimate losses should be compared with gross cumulative paid losses; a net ultimate estimate should be compared with net paid losses. A difference between gross and net may reflect recoveries or other adjustments rather than unpaid claims.

The same consistency requirement applies to the cohort and timing definitions. If an ultimate estimate covers accident-year claims, subtract paid losses for that accident year—not payments for a policy-year grouping or a different set of claims. If the paid figure is cumulative through the valuation date, do not subtract only the latest period's payments. All amounts should also use a consistent treatment of loss adjustments and monetary units.

Worked Example: Converting a Selected Ultimate Multiplier to Unpaid Loss

A hypothetical equipment insurer has cumulative paid losses of $640,000 for a cohort at the valuation date. A previously selected ultimate multiplier is 1.75, applied to that cumulative paid amount to obtain the estimated ultimate. Calculate the estimated ultimate and unpaid liability. Assume both amounts are undiscounted gross losses and all figures are in dollars.

Step 1: Estimate ultimate loss. Apply the selected multiplier to cumulative paid loss:

\hat U=\$640,000\times1.75=\$1,120,000.

Step 2: Subtract paid loss. The estimated unpaid liability is $ \$1,120,000-\$640,000=\$480,000$.

Step 3: Check the decomposition. The estimated ultimate equals paid plus unpaid: $ \$640,000+\$480,000=\$1,120,000$.

Method and trap avoided. First use the multiplier to estimate total ultimate loss, then subtract the current cumulative paid amount. Treating the projected ultimate of $1,120,000 as the liability would count already-paid losses again.

Common Exam Traps and Interpretation

The calculation is often short, so errors tend to arise from interpreting the quantities rather than doing the subtraction. Keep the question's requested measure in view: “ultimate loss” and “unpaid liability” are not interchangeable.

Reporting ultimate as unpaid. If estimated ultimate is $900,000 and cumulative paid is $540,000, the unpaid estimate is $ \$900,000-\$540,000=\$360,000$, not $900,000.
Subtracting incremental rather than cumulative payments. The formula uses all payments through the valuation date. If $120,000 was paid in the latest period but cumulative paid is $540,000, subtract $540,000.
Mixing bases. Do not subtract net paid from gross ultimate, or vice versa, without first bringing the amounts to a common basis.
Assuming the estimate must fall by the amount paid. That conclusion holds only if the ultimate estimate is unchanged. When the ultimate estimate is revised, use the full change formula.
Ignoring a negative result. If $\hat U<C(t)$, the direct calculation gives a negative indicated amount. Do not silently replace it with zero unless the problem explicitly requires a floor. Instead, check the cohort, paid-loss basis, valuation date, and selected ultimate estimate; the result may also reflect a genuine estimate revision or adjustment.

The liability calculated here is limited to the unpaid loss measure represented by the selected ultimate and cumulative paid amounts. Other claim-related amounts, such as expenses or separately measured obligations, should not be assumed to be included unless the data and question define the ultimate and paid figures to include them.

Key takeaway: At a valuation date, estimated unpaid loss is estimated ultimate loss minus cumulative paid loss through that date. Apply the subtraction cohort by cohort, keep the loss basis consistent, and distinguish the remaining liability from the total ultimate estimate.

Check Your Understanding

For each problem, calculate the requested unpaid liability on the stated basis.

A hypothetical cohort has estimated ultimate losses of $725,000 and cumulative paid losses of $410,000. What is the estimated unpaid liability? (Answer: $315,000)
Three hypothetical cohorts have estimated ultimate losses of $1,200,000, $900,000, and $650,000, with cumulative paid losses of $840,000, $540,000, and $260,000, respectively. What is the total estimated unpaid liability? (Answer: $1,110,000)
A hypothetical cohort has cumulative paid loss of $880,000 and a selected ultimate multiplier of 1.45. Apply the multiplier to paid loss to estimate ultimate, then calculate unpaid loss. (Answer: Estimated ultimate $1,276,000; unpaid liability $396,000)
At the first valuation date, a hypothetical cohort has estimated ultimate loss of $2,400,000 and cumulative paid loss of $1,500,000. At the next date, an additional $250,000 has been paid and the revised ultimate estimate is $2,550,000. What is the new unpaid estimate and how much did the liability change? (Answer: New unpaid liability $800,000; change $-\$100,000$)
On a consistent hypothetical net basis, estimated ultimate loss is $3,240,000 and cumulative paid loss is $2,000,000. Separately, gross ultimate loss is $3,400,000 and gross paid loss is $2,100,000. Calculate both the net and gross unpaid liabilities, and state their difference. (Answer: Net $1,240,000; gross $1,300,000; gross exceeds net by $60,000)

What You'll Learn