Calendar-Time Changes in Claim Handling

Why Calendar Time Matters

Claim-handling practices can change at a particular point in calendar time: an insurer might introduce a new settlement authority, change how quickly adjusters review files, or implement a new case-reserving guideline. Such a change may affect claims from several accident years at once, because claims from different accident years can all be handled during the same calendar period.

This creates a distinct diagnostic question in Berquist-Sherman work: do observed changes follow accident-year maturity, or do they appear across several accident years at a common valuation date? A calendar-time pattern can signal an operational influence on paid or incurred development. It does not, by itself, establish the cause or quantify an adjustment.

Definition: In an annual loss development triangle, the calendar year associated with accident year $AY$ and development age $a$ is $$ CY=AY+a-1. $$ Equivalently, the accident year on calendar-year diagonal $CY$ at development age $a$ is $$ AY=CY-a+1. $$ A calendar-year diagonal is the set of triangle cells observed at the same calendar-year valuation.

The convention assumes accident years and development ages are expressed in annual periods, with age 1 corresponding to the accident year. For example, the age-1 cell for accident year 2025 is in calendar year 2025; its age-2 cell is in calendar year 2026. In quarterly or irregularly valued data, the periods must be mapped using the actual valuation dates rather than this annual formula.

Mapping a Handling Change to Triangle Cells

A change effective in calendar year $T$ can influence the claims being handled in that year, whatever their accident year or development age. In a triangle, those cells lie on the $T$ diagonal. As claims continue to develop, later cells for those same accident years occur in later calendar years; the original change is not automatically a permanent change to every later diagonal.

This distinction matters when interpreting adjusted triangles. A faster payment process might alter paid amounts or settlement rates on the implementation-year diagonal. A change to case-reserving practice might alter incurred amounts and case outstanding on that diagonal. These effects may produce different patterns in paid and incurred data, and the observed pattern can also reflect changes in claim mix, coverage, or other conditions.

Worked Example: Locating the Cells Around an Implementation Date

A hypothetical insurer introduces a revised claim-review process at the start of calendar year 2025. Identify the age-1, age-2, and age-3 cells on the 2025 calendar diagonal. Then identify the calendar years for accident year 2025 at ages 1, 2, and 3.

Method. Use $CY=AY+a-1$ to find calendar year from accident year and age. Use $AY=CY-a+1$ to find the accident year on a stated calendar diagonal.

On the 2025 diagonal:

At age 1, $AY=2025-1+1=2025$, so the cell is $(2025,1)$. At age 2, $AY=2025-2+1=2024$, giving $(2024,2)$. At age 3, $AY=2025-3+1=2023$, giving $(2023,3)$.

For accident year 2025, age 1 is in $2025+1-1=2025$. Age 2 is in $2025+2-1=2026$. Age 3 is in $2025+3-1=2027$.

Check. The three cells on the 2025 diagonal all satisfy $AY+a-1=2025$: $2025+1-1=2025$, $2024+2-1=2025$, and $2023+3-1=2025$. In particular, accident year 2025 at age 2 is in calendar year 2026, not 2025. Common mistake avoided: confusing the age-2 cell of accident year 2025 with the age-2 cell on the 2025 diagonal.

Reading Valuation-Date Patterns

A calendar-time change is most readily suspected when a measure changes across several accident years at about the same valuation date, rather than changing only as each accident year reaches a particular age. Useful measures depend on the handling change and the data available. Examples include paid-to-incurred ratios, closure rates, average case outstanding per open claim, or changes in paid development.

When comparing such measures, identify both the calendar period and the development age for every cell. Cells on one calendar diagonal have different ages, so their levels need not be equal even if a common operational change affected them. The diagnostic question is whether the pattern changes in a coordinated way relative to appropriate maturity-specific expectations.

Worked Example: A Shared Change in Paid-to-Incurred Ratios

A hypothetical insurer calculates paid-to-incurred ratios for selected accident years. At the end of calendar year 2024, the ratios are 0.62 for accident year 2022 at age 3, 0.61 for accident year 2023 at age 2, and 0.60 for accident year 2024 at age 1. At the end of calendar year 2025, the corresponding ratios are 0.72, 0.71, and 0.70, respectively, at ages 4, 3, and 2. A new payment workflow began during 2025. Calculate the changes and assess what the pattern suggests.

Method. Compare each accident year with itself across the two valuation dates; do not compare ratios across accident years as if they had the same maturity.

For accident year 2022, the ratio changes from 0.62 to 0.72, an increase of $0.72-0.62=0.10$, or 10 percentage points. For accident year 2023, the increase is $0.71-0.61=0.10$, also 10 percentage points. For accident year 2024, it is $0.70-0.60=0.10$, again 10 percentage points.

The three cohorts all show the same increase between the 2024 and 2025 valuations, although they are at different ages. That coordinated movement is consistent with a calendar-time influence and makes the 2025 workflow change a reasonable hypothesis to investigate.

Check. Each comparison is within one accident year: the cells move from age 3 to 4, age 2 to 3, or age 1 to 2, respectively, as calendar time advances from 2024 to 2025. The ratios alone do not prove that the workflow caused the movement; the incurred estimates, claim mix, and other changes could also affect them. Common mistake avoided: treating a shared pattern as causal proof, or comparing different maturity levels without accounting for age.

The valuation date identifies when the data were observed; it is not itself a measure of development age. A displayed triangle may therefore show a handling change as a diagonal pattern even when the underlying accident years have different amounts and maturities. This is why the diagonal mapping is a necessary first step before interpreting an adjusted triangle.

How a Calendar Shift Can Change Apparent Development

Link ratios compare cumulative amounts at successive development ages. A change in handling during the calendar period between those observations can influence a link ratio. The following calculation illustrates the diagnostic comparison: each observed link ratio is compared with a stated maturity-specific benchmark. The benchmark is assumed for illustration, not inferred from the cells.

Diagnostic calculation: For cumulative amount $C$ at the earlier age and link ratio $f$, the benchmark cumulative amount at the later age is $C f$. The difference between the observed later cumulative amount and $C f$ shows the departure from that benchmark. A departure is not, on its own, an estimate of the causal impact of a handling change.

Worked Example: Comparing Link Ratios Across a Common Calendar Period

A hypothetical handling change begins in calendar year 2025. For three accident years, the cumulative paid losses at the prior valuation and the observed next-age amounts are: accident year 2022, 1,000 at age 3 and 1,250 at age 4; accident year 2023, 800 at age 2 and 1,160 at age 3; accident year 2024, 600 at age 1 and 960 at age 2. Maturity-specific benchmark link ratios are 1.20, 1.35, and 1.50, respectively. Compare the observed amounts with the benchmarks.

Method. Divide each observed later amount by its earlier amount to calculate the observed link ratio. Multiply the earlier amount by the benchmark link ratio to calculate the benchmark later amount.

For accident year 2022, the observed link ratio is $1{,}250/1{,}000=1.25$. The benchmark amount is $1{,}000(1.20)=1{,}200$, so the observed amount is $1{,}250-1{,}200=50$ above the benchmark.

For accident year 2023, the observed link ratio is $1{,}160/800=1.45$. The benchmark amount is $800(1.35)=1{,}080$, so the difference is $1{,}160-1{,}080=80$.

For accident year 2024, the observed link ratio is $960/600=1.60$. The benchmark amount is $600(1.50)=900$, so the difference is $960-900=60$.

All three later observations occur in calendar year 2025: accident year 2022 at age 4, accident year 2023 at age 3, and accident year 2024 at age 2. Each observed link ratio exceeds its stated benchmark, and the total of the three benchmark departures is $50+80+60=190$.

Check. The amounts implied by the benchmarks are 1,200, 1,080, and 900; adding the respective departures gives the observed 1,250, 1,160, and 960. The common calendar period makes the pattern worth investigating, but the 190 is not automatically an adjustment to ultimate losses. The benchmark assumptions, mix, and other influences still require consideration. Common mistake avoided: assuming that the sum of departures is a causal handling adjustment merely because the cells share a calendar year.

A Practical Diagnostic Sequence

Pin down the change date.
Record the effective calendar period and the type of operational change. A general statement such as “handling became faster” is not enough to specify which measures should move.

Map the diagonal.
Use $CY=AY+a-1$ to identify the cells exposed to the change. Check the actual valuation dates if the data are not annual.

Inspect relevant measures.
Look at measures that could respond to the stated change, such as payment development for a payment-process change or case estimates for a reserving-guideline change.

Compare at appropriate maturities.
Use maturity-specific expectations or within-accident-year comparisons. Do not interpret different ages on the same diagonal as though they were directly interchangeable.

Describe the evidence cautiously.
State whether the pattern is consistent with a calendar-time influence. Separate that observation from a claim that the change caused it or from a proposed numerical adjustment.

Calendar-time diagnosis is useful because an aggregate triangle can blend a genuine operational shift with the normal development pattern. If several cohorts change around the same valuation date, the analyst has a reason to inspect the handling history and the relevant measures before selecting or applying development assumptions. The interpretation remains a hypothesis to test against the data and the known circumstances.

Common Exam Traps

Using the wrong direction of the mapping. $AY+a-1$ gives the calendar year of a cell. To find the accident year on a known diagonal, use $CY-a+1$.
Forgetting the one-period offset. Age 1 is in the accident year, so calendar year is $AY+a-1$, not $AY+a$.
Calling a calendar diagonal an age column. A diagonal holds cells from several accident years at different ages. An age column holds one age across accident years and therefore spans calendar years.
Overstating what a pattern proves. A coordinated change can support an operational hypothesis, but it does not eliminate claim-mix or other explanations.

Key takeaway: A calendar-time handling change can affect cells from multiple accident years along one triangle diagonal. Map the cells correctly, compare relevant measures at suitable maturities, and treat the resulting pattern as diagnostic evidence rather than automatic proof or an adjustment formula. This topic is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Check Your Understanding

For each problem, map the calendar period carefully before interpreting the triangle.

A hypothetical claim-review change begins in calendar year 2027. List the cells on the 2027 diagonal at ages 1, 2, and 3. (Answer: $(2027,1)$, $(2026,2)$, and $(2025,3)$.)
What calendar year contains the age-3 cell for accident year 2023? (Answer: $2023+3-1=2025$.)
A ratio rises from 0.58 at one valuation to 0.67 at the next valuation for the same accident year. Give the change in percentage points and the relative increase from the original ratio, rounded to 1 decimal place. (Answer: 9 percentage points; $0.09/0.58=15.5\%$.)
For one accident year, cumulative paid loss is 900 at the earlier age and 1,260 at the next age. A maturity-specific benchmark link ratio is 1.30. Calculate the observed link ratio, benchmark later amount, and departure from benchmark. (Answer: observed ratio $1{,}260/900=1.40$; benchmark amount $900(1.30)=1{,}170$; departure $1{,}260-1{,}170=90$.)
Three accident years have paid-to-incurred ratios of 0.64, 0.65, and 0.63 at one valuation, and 0.74, 0.75, and 0.73 at the next. Calculate the mean ratio at each valuation and the increase. What is a cautious interpretation if the later valuation follows a handling change? (Answer: means are $0.64$ and $0.74$, an increase of $0.10$, or 10 percentage points; the coordinated pattern is consistent with a calendar-time influence but does not prove causation.)

What You'll Learn