Accident Period, Development Age, and Calendar Period

Why Three Time Labels Matter

A loss development triangle organizes observations by more than one clock. The row identifies when the underlying accidents or exposures originated, the column identifies how long they have developed, and the diagonal identifies the calendar period associated with those observations. These labels answer different questions, even when they all use years as their units.

Keeping the three periods distinct matters when interpreting a pattern in a triangle. A change down a row concerns different accident periods; a change across a row concerns different development ages; a change along a diagonal concerns different calendar periods. Confusing them can lead to attributing a change to the wrong source.

In a conventional annual triangle, accident years label the rows and development ages are numbered 1, 2, 3, and so on. Development age 1 represents the first annual development interval for an accident year. Thus, a cell in accident year 2021 at development age 1 is associated with calendar year 2021; its development-age-2 cell is associated with calendar year 2022.

This formula describes the usual annual indexing convention, not every possible data layout. Always check what a data source calls development age 1. If a source instead labels columns by elapsed periods starting at age 0, the index formula changes. The calendar period assigned to a cell also depends on how the triangle defines its observation or valuation date.

Reading Rows, Columns, and Diagonals

A row holds one accident period fixed while development age increases. Moving across that row therefore advances the development clock. A column holds development age fixed while the accident period changes. Moving down that column changes the origin cohort. A diagonal running from the upper left toward the lower right holds the calendar period fixed.

For example, the cells for accident year 2020 at development age 3 and accident year 2021 at development age 2 share calendar year 2022. They lie on the same diagonal because both satisfy the same calendar-year mapping. The observations may refer to different accident cohorts and different maturity levels, but they are associated with the same calendar period.

Why the Diagonal Mapping Holds

The annual formula follows directly from counting the calendar-year steps after the accident year. Development age 1 has zero calendar-year steps after the origin year. Development age 2 has one step; development age 3 has two. In general, development age \(j\) is \(j-1\) steps after the origin year.

To see why diagonals share a calendar period, take any two cells with indices \((i_1,j_1)\) and \((i_2,j_2)\). They have the same calendar period exactly when $$ i_1+j_1-1=i_2+j_2-1, $$ which is equivalent to \(i_1+j_1=i_2+j_2\). As the accident-year index increases by one down the triangle, the development-age index must decrease by one to keep the sum unchanged. That is the diagonal direction.

This is a bookkeeping identity, not a claim that the cells on a diagonal contain identical losses or measure the same accident cohort. They concern different origin periods and different ages. The identity only says that their calendar-period index is the same.

Quarterly and Other Periodic Triangles

The same logic applies when the periods are quarters, months, or another consistent unit. Use an index that advances by one for each period, rather than adding calendar labels as though every step were a year. In a quarterly triangle, for instance, the development-age-1 cell is associated with the origin quarter, and each increase of one development age advances one quarter.

A practical way to calculate is to convert the origin quarter into a sequential quarter index, add \(j-1\), and convert back to a year and quarter. This avoids errors when adding across a year boundary.

For monthly data, the identical procedure uses months as the unit: advance \(j-1\) months from the origin month. The formula is about equally spaced periods. It does not mean that every development interval must be a calendar year; the unit of the index must match the triangle’s construction.

Calendar Period Is Not the Same as Accident Period

Accident period describes the origin of the claims, while calendar period describes when the cell’s recorded experience or valuation belongs. For example, a claim arising from an accident in 2021 may produce observations in later calendar years as it develops. The accident year remains 2021; the calendar period changes as the claim is observed at later development ages.

This distinction is useful when examining possible calendar-period influences, such as a change in claims handling, a benefit rule, or an external environment. A diagonal organizes cells associated with the same calendar period, so it can help identify where such a common-period effect might appear. A diagonal pattern alone does not establish that a calendar-period influence caused the observed values; interpretation requires attention to what the data represent and what else changed.

The calendar period assigned to a cell must follow the triangle’s reporting convention. Some data describe amounts accumulated or recorded during an interval; other data are measured at a valuation date. The period index locates the cell in calendar time, but does not by itself specify whether the underlying quantity is a period amount, a value-to-date, or a snapshot. In particular, the period mapping does not determine whether the triangle is incremental or cumulative.

Common Indexing Traps

• Adding the full development age. With age 1 in the accident year, age \(j\) is \(j-1\) periods after origin, not \(j\) periods after origin.
• Treating a diagonal as one accident cohort. A diagonal holds calendar period fixed; its cells generally have different accident periods and development ages.
• Mixing time units. A quarterly age must be counted in quarters, not years. Convert to a sequential index when the calendar label crosses a year boundary.
• Assuming every source uses age 1 in the origin period. Check the stated column convention. If the first column is elapsed age 0, adjust the mapping.
• Inferring the quantity from the time label. A calendar-year label alone does not say whether the cell is a period amount or a valuation snapshot.