Building a Paid Loss Triangle from Claim Transactions

From Claim Transactions to a Paid Triangle

A claim-payment ledger records individual cash movements, while a loss triangle organizes those movements by accident period and development age. Constructing the triangle is therefore a classification and aggregation task: each transaction must be assigned to the period in which the claim arose and the development interval in which the payment was recorded.

This tutorial constructs an incremental paid loss triangle. Each cell contains the net amount paid during one development interval for claims from one accident period. The result depends on consistent definitions of accident period, valuation date, and transaction date. A transaction assigned to the wrong interval can distort the triangle even when the overall ledger total is correct.

Definition: Let $i$ identify an accident period and $j$ identify a development interval. The incremental paid amount $I_{i,j}$ is the sum of the signed paid transaction amounts assigned to accident period $i$ and development interval $j$. A transaction amount is signed when the accounting record includes a reversal or other adjustment that increases or decreases net paid amounts.

Set the Periods Before Adding Amounts

An accident period groups claims according to when the insured events occurred. Depending on the data, it may be an accident year, quarter, or month. A development interval identifies how far a payment falls after the start of its accident period. For annual accident periods, development age 1 is the accident year itself, development age 2 is the following calendar year, and so on.

For annual periods, if $i$ denotes the accident year and $k$ the calendar year containing the payment, then the payment's development age is $j=k-i+1$. For example, a payment in calendar year 2024 for an accident-year 2023 claim is at development age 2. The same calendar year can contain transactions at different development ages for different accident periods.

In practice, period boundaries must be specified precisely. For quarters, for example, a transaction dated on June 30 belongs to the second calendar quarter, while one dated July 1 belongs to the third. Use the date field and accounting convention specified for the data: a posting date and a payment-effective date need not be identical. Do not mix them within one triangle.

Construction rule: For each transaction, identify its accident period, determine its development interval from the chosen date convention, and add its signed paid amount to the corresponding cell. Include only transaction types that match the stated paid-loss basis; reserve changes are not paid transactions.

Worked Construction: Annual Accident Periods

The following hypothetical ledger contains paid transactions for accident years 2022–2024. Amounts are in dollars. Valuations are at each calendar year-end, and development age 1 is the accident year itself.

Worked Example: Aggregating an Annual Payment Ledger

Construct the incremental paid triangle from these transactions, then check its total against the ledger.

Accident year	Payment calendar year	Transaction amounts ($)
2022	2022	80, 35
2022	2023	40, 25
2022	2024	20
2023	2023	90, 30
2023	2024	45
2024	2024	110, 15

Step 1: Group each row of transactions. For accident year 2022, the calendar-year subtotals are $80+35=115$ in 2022, $40+25=65$ in 2023, and $20$ in 2024. These correspond to development ages 1, 2, and 3.

For accident year 2023, the 2023 subtotal is $90+30=120$ at age 1, and the 2024 subtotal is $45$ at age 2. For accident year 2024, the 2024 subtotal is $110+15=125$ at age 1.

Step 2: Place the subtotals in the triangle. Each amount stays with its accident year but is placed in the column for its development age.

Accident year	Age 1	Age 2	Age 3
2022	115	65	20
2023	120	45
2024	125

Step 3: Reconcile. The accident-year subtotals are $115+65+20=200$, $120+45=165$, and $125$. Their total is $200+165+125=490$. Summing the original transaction amounts gives the same result: $80+35+40+25+20+90+30+45+110+15=490$. The construction reconciles.

Method and mistake avoided. This is transaction-level grouping by accident year and development age. The common mistake is to assign all 2023 payments to development age 1. The 2023 payments for accident year 2022 are at age 2, not age 1.

Development Age Within a Calendar Quarter

The same method applies when accident periods are shorter than a year. The key is to define age 1 as the accident quarter, then advance one development interval for each subsequent quarter. Payments are allocated according to the quarter in which the selected transaction date falls.

Worked Example: Quarterly Boundaries and Different Ages

A hypothetical insurer groups claims by accident quarter. A development interval is one calendar quarter, with age 1 equal to the accident quarter. Use the following payment records, valued through September 30, 2025, to construct the observed incremental cells.

Accident quarter	Payment date	Amount ($)
2025 Q1	March 31, 2025	70
2025 Q1	April 1, 2025	18
2025 Q1	June 30, 2025	12
2025 Q1	July 2, 2025	9
2025 Q2	April 15, 2025	55
2025 Q2	June 30, 2025	5
2025 Q2	July 10, 2025	14

Step 1: Assign the dates to calendar quarters. March 31 is in Q1; April 1 and June 30 are in Q2; July 2 and July 10 are in Q3.

Step 2: Convert calendar quarter to development age. For accident quarter 2025 Q1, Q1 is age 1, Q2 is age 2, and Q3 is age 3. Thus its three incremental amounts are $70$, $18+12=30$, and $9$. For accident quarter 2025 Q2, Q2 is age 1 and Q3 is age 2, giving $55+5=60$ and $14$.

Step 3: Check totals. The Q1 accident-period total is $70+30+9=109$, matching its transactions $70+18+12+9=109$. The Q2 accident-period total is $60+14=74$, matching $55+5+14=74$. Across both accident periods, the total is $109+74=183$.

Accident quarter	Age 1	Age 2	Age 3
2025 Q1	70	30	9
2025 Q2	60	14

Method and mistake avoided. The method maps dates to calendar quarters first, then maps each calendar quarter to an age separately for each accident quarter. A common error is to treat the whole calendar-quarter total as one development-age amount. For instance, the Q2 calendar-quarter total here is $18+12+55+5=90$, but it comprises age 2 for 2025 Q1 and age 1 for 2025 Q2.

Why Incremental Cells Reconcile to Paid Amounts

The construction works because the development intervals partition the payment records for an accident period: each included transaction belongs to one interval and is counted once. Adding cells across development ages therefore recovers the total signed amount recorded for that accident period through the valuation date.

I_{i,j}=\sum_{r:\,a_r=i,\ d_r=j} z_r \qquad\text{and}\qquad C_{i,m}=\sum_{j=1}^{m} I_{i,j}

Here $z_r$ is the signed amount of transaction $r$, $a_r$ is its assigned accident period, $d_r$ is its development interval, and $C_{i,m}$ is the cumulative paid amount for accident period $i$ through age $m$. The first equation aggregates transactions into incremental cells. The second adds those cells from age 1 through age $m$.

To justify the second equation, substitute the cell definition into the sum. Each transaction with accident period $i$ and development age no later than $m$ appears in exactly one of the cell sums, so the result is precisely the total paid amount through age $m$. This is why cumulative paid amounts can be derived from incremental cells, but should not be placed in an incremental triangle.

For example, the 2022 row in the first example has incremental amounts $115,65,20$. Its cumulative amounts are $115$, $115+65=180$, and $180+20=200$. The cumulative row is $115,180,200$; the incremental row remains $115,65,20$.

Signed Adjustments and Reconciliation Checks

Some ledgers contain negative entries, such as reversals of previously recorded payments. Whether recoveries or other adjustments are included depends on the stated paid-loss basis and the source-data convention. Apply that convention consistently. If a reversal is recorded in a later period, allocate it using the transaction date or other date basis specified for the triangle; do not silently move it to the original payment period.

The following example shows how a signed reversal affects the cell total.

Worked Example: A Reversal in the Payment Ledger

A hypothetical annual ledger, in dollars, contains these records: accident year 2023 has payments of $90$ and $60$, followed by a $-15$ reversal, all recorded in calendar year 2023; it also has a $25$ payment recorded in 2024. Accident year 2024 has payments of $80$ and $20$ in 2024 and a $30$ payment in 2025. Construct the incremental cells through 2025 and reconcile.

Step 1: Aggregate the 2023 accident-year transactions. At age 1, the net amount is $90+60-15=135$. At age 2, the 2024 amount is $25$. Its total through 2024 is $135+25=160$.

Step 2: Aggregate the 2024 accident-year transactions. At age 1, the 2024 amount is $80+20=100$. At age 2, the 2025 amount is $30$. Its total through 2025 is $100+30=130$.

Accident year	Age 1	Age 2
2023	135	25
2024	100	30

Step 3: Check the ledger total. The triangle total is $135+25+100+30=290$. The transaction total is $90+60-15+25+80+20+30=290$, so the figures reconcile.

Method and mistake avoided. The signed transaction amounts are aggregated within their assigned periods. Omitting the reversal would overstate the 2023 age-1 cell by $15$, and the overall triangle by $15$. The reversal is not automatically reallocated to the date of the original payment.

Common Construction Traps

Mixing accident period and calendar period. A calendar-year subtotal can contain payments at several development ages. Determine age separately for each accident period.
Using cumulative amounts in incremental cells. Cumulative paid through age 2 includes age 1; an incremental age-2 cell includes only transactions assigned to age 2.
Ignoring date boundaries. A payment dated on the first day of a new quarter belongs to that quarter under calendar-quarter grouping, not the prior one.
Dropping or misplacing signed adjustments. Follow the stated paid basis and transaction-date convention, and retain the sign when aggregating.
Forgetting the reconciliation. If the triangle and included ledger transactions do not have the same total, check for omitted records, duplicate records, incorrect accident-period assignments, and inconsistent valuation cutoffs.

Key takeaway: Define the accident periods, development intervals, transaction-date convention, and paid-loss basis first. Then classify every transaction once, sum signed amounts within cells, and reconcile the triangle total to the included ledger.

Check Your Understanding

For each problem, construct or identify the requested incremental paid amounts. Assume all listed records are valid paid transactions unless a negative amount is shown.

A ledger has accident-year 2024 payments of $30$ and $15$ in 2024 and $25$ in 2025. Accident-year 2025 has payments of $40$ and $10$ in 2025. What are the incremental cells for each accident year? (Answer: 2024: age 1 = 45, age 2 = 25; 2025: age 1 = 50.)
For accident year 2022, payments are $120$ in 2022, $35$ in 2023, and $15$ in 2024. State the incremental row and cumulative paid amount through age 2. (Answer: Incremental row = 120, 35, 15; cumulative through age 2 = 155.)
Accident periods are calendar quarters, with age 1 equal to the accident quarter. For accident quarter 2025 Q4, payments are $40$ on December 20, 2025; $12$ on January 1, 2026; $8$ on March 31, 2026; and $5$ on April 1, 2026. Give the incremental amounts through age 3. (Answer: Age 1 = 40; age 2 = 20; age 3 = 5.)
Accident-year 2023 has transactions of $90$ and $-15$ in 2023 and $25$ in 2024. Give the incremental cells and total paid amount through age 2. (Answer: Age 1 = 75; age 2 = 25; total through age 2 = 100.)
Accident-year 2024 has $50$ paid in calendar year 2024 and $20$ in 2025. Accident-year 2025 has $60$ paid in calendar year 2025. Give the three observed incremental cells and the total paid in calendar year 2025 across both accident years. (Answer: 2024 age 1 = 50, 2024 age 2 = 20, 2025 age 1 = 60; calendar-year 2025 total = 80.)

What You'll Learn