Why the Valuation Date Matters
A claim record is not a timeless fact. At one date, a claim may not yet have been reported; at another, it may be present, and the amounts recorded for it may have changed. A ratemaking data set therefore needs a clear answer to the question: what was known, and as of when? Without that answer, two experience periods can look different simply because they were observed at different stages of claim development.
The valuation date is the cutoff that defines the claim information available for an analysis. It is separate from the dates that define the experience period. Consistent cutoffs are essential when assembling claim records or comparing summaries. However, a common cutoff does not make all periods equally mature: a period ending earlier has had more time for claims to be reported and recorded.
An experience period might be defined by occurrence dates, policy dates, or another exposure basis. The valuation date does not change that definition. It determines which claim information is available for the chosen period. The earlier material on exposure bases and claim-data construction establishes why the period definition and its exposure denominator must be specified; here, the focus is the cutoff on claim information.
Measuring Maturity
For an occurrence-based period, a practical maturity measure is the time from the period end to the valuation date. It describes how long the period has had to generate reported claims. If a period ends on December 31 and the valuation date is June 30 of the following year, its age is six months. For individual claims, elapsed time from occurrence to valuation can also help explain why some claims have had more opportunity to be reported than others.
Maturity is not the same as completeness. A six-month-old period may still receive reports, and two periods of the same age may have different reporting patterns. Unless a specified operational rule defines a period as sufficiently mature, age is a warning and comparison tool, not proof that all claims are known.
The subtraction is calendar elapsed time, not a count of exposure units. State the date convention used, especially when dates fall in the middle of a month. A summary that reports ages in whole months should say so; it should not imply day-level precision that its convention does not support.
Worked Example: Which Claims Are Visible at Each Cutoff?
A hypothetical insurer is assembling occurrence-year 2024 claim records. The three claims below are the only claims in the exercise. Determine which are observable at January 1, 2025 and July 1, 2025, and calculate the age of each included claim in whole months. Assume the month-age convention counts whole calendar months from the occurrence date to the valuation date. No interest or mortality assumption is relevant; these are deterministic claim records, and no probability distribution is assumed.
| Claim | Occurrence date | Report date |
|---|---|---|
| A | January 1, 2024 | January 10, 2024 |
| B | July 1, 2024 | August 1, 2024 |
| C | December 1, 2024 | February 1, 2025 |
Method. Include a claim only when its report date is on or before the valuation date. Then calculate elapsed whole months from occurrence to valuation for each included claim.
At January 1, 2025: Claim A was reported by the cutoff, as was claim B. Claim C was not reported until February 1, 2025, so it is not in the January extract. Claim A is 12 months old: January 1, 2024 to January 1, 2025. Claim B is 6 months old: July 1, 2024 to January 1, 2025.
At July 1, 2025: All three claims have been reported by the cutoff. Their ages are 18 months for A, 12 months for B, and 7 months for C. For example, December 1, 2024 to July 1, 2025 is seven whole calendar months.
Check. The January snapshot contains 2 of the 3 claims listed; the July snapshot contains all 3. The change is due to the later report of C, not a change in the definition of occurrence-year 2024. The common mistake is to include C in the January snapshot because it occurred during 2024; the correction is to apply the report-date cutoff as well as the occurrence-period rule.
Use One Valuation Date for a Comparison
When multiple experience periods are extracted, use the same valuation date for every period and for every claim measure intended to be compared. The same principle applies when paid and incurred fields are both present: each field must be read from a clearly identified snapshot, and the snapshots should share a common cutoff for a like-for-like comparison. A common valuation date does not mean paid and incurred amounts are the same measure; it means their recorded values are taken as of the same date. The later material on paid, incurred, and ultimate loss bases treats the differences between those measures.
The distinction between a common cutoff and common maturity is important. A common cutoff makes the information date consistent, but older experience periods have had longer to develop by that date. Conversely, observing each period at the same age can mean using different calendar valuation dates. A comparison should state which of these features it is controlling and why.
Worked Example: Aligning Paid and Incurred Snapshots
A hypothetical portfolio has two occurrence-year summaries, with amounts in thousands of dollars. A data team wants to compare the years using a June 30, 2025 valuation date. The extract contains the following values. No interest or mortality assumption is relevant; the amounts are deterministic snapshot values, with no probability distribution assumed.
| Occurrence year | Valuation date | Paid amount ($000) | Incurred amount ($000) |
|---|---|---|---|
| 2023 | June 30, 2025 | 100 | 140 |
| 2023 | December 31, 2025 | 108 | 150 |
| 2024 | June 30, 2025 | 60 | 90 |
Method. Select the rows whose valuation date equals the stated cutoff, then compare like fields across occurrence years.
For June 30, 2025, the 2023 paid value is $100,000 and its incurred value is $140,000. The 2024 paid value is $60,000 and its incurred value is $90,000. The sums across these two years are $160,000 paid and $230,000 incurred: $100,000 + $60,000 = $160,000, and $140,000 + $90,000 = $230,000.
A mismatched-date calculation would use the December 31, 2025 values for 2023 and the June 30, 2025 values for 2024. It would produce paid of $108,000 + $60,000 = $168,000 and incurred of $150,000 + $90,000 = $240,000. Those totals mix valuation dates and do not describe one consistent snapshot. The corrected totals at the requested common date are $160,000 and $230,000.
Maturity check. At June 30, 2025, occurrence year 2023 has ended 18 months earlier, while occurrence year 2024 has ended 6 months earlier. Thus the corrected amounts share a valuation date, but the periods do not have equal maturity. The common mistake is to treat matching valuation dates as proof of matching completeness; the correction is to report period age separately.
Why Recent Periods Are Incomplete
A reported-claim count illustrates the basic mechanism. Fix an occurrence period and let \(R(t)\) be the set of claims from that period reported by elapsed time \(t\) after the period ends. If no records are removed under the stated data rules, then for \(t_1 \leq t_2\),
The justification is direct: a claim reported by \(t_1\) has also been reported by the later time \(t_2\). The later information set contains the earlier reported claims and may contain additional claims reported in the interval. Consequently, a recently ended period can have a lower observed count merely because less time has elapsed. Recorded amounts can also change between snapshots, so a later extract may differ even for claims already present.
This reasoning does not tell an analyst how many unreported claims exist or what the eventual amount will be. A valuation date identifies the observed data; it does not, by itself, estimate missing information. Do not turn an observed count or amount into an ultimate estimate without an appropriate method and assumptions. For this topic, the key task is to flag the limitation, keep dates explicit, and avoid interpreting an immature observation as a fully comparable result.
Worked Example: Screening Quarterly Experience for Immaturity
A hypothetical insurer reviews claims from three 2025 occurrence quarters at a September 30, 2025 valuation date. Each quarter has 100 policy-years of exposure. The claim counts below are determined from report dates; each quarter has three claims in the complete chronology. Use a six-month period-age rule only as a screening rule: periods younger than six months are flagged as immature. No interest or mortality assumption is relevant; the records are deterministic, and no probability distribution is assumed.
| Occurrence quarter | Quarter end | Claim report dates |
|---|---|---|
| Q1 | March 31, 2025 | January 20, May 15, October 10 |
| Q2 | June 30, 2025 | April 20, June 1, October 20 |
| Q3 | September 30, 2025 | August 1, October 5, October 10 |
Step 1: Calculate period ages. At September 30, Q1 ended six months earlier, Q2 ended three months earlier, and Q3 ended on the valuation date, so their ages are 6, 3, and 0 months.
Step 2: Count reports by the cutoff. In Q1, January 20 and May 15 are on or before September 30, so 2 claims are included; October 10 is after the cutoff. In Q2, April 20 and June 1 are included, so the observed count is 2; October 20 is after the cutoff. In Q3, August 1 is included, while October 5 and October 10 are after the cutoff, so the observed count is 1.
Step 3: Apply the stated screening rule. Q1 meets the six-month age threshold. Q2 and Q3 do not, so they are flagged as immature. With 100 policy-years in each quarter, the observed counts per policy-year are \(2/100 = 0.020\), \(2/100 = 0.020\), and \(1/100 = 0.010\), respectively.
Check and limitation. The September 30 snapshot contains 5 reports across the quarters. The later report dates in the table show that the chronology eventually contains 9 claims, but those future reports were not available at the September 30 valuation date. The observed rates are therefore not evidence that Q3 necessarily has a lower underlying claim frequency. The common mistake is to read the smallest observed rate as the best-performing quarter; the correction is to note that Q3 is age zero and its observed claims are incomplete.
Common Valuation-Date Traps
- Using occurrence date as the reporting cutoff. A claim can belong to an experience period by occurrence date and still be absent from an extract if it was not reported by the valuation date.
- Mixing snapshot dates. Paid or incurred fields from different valuation dates do not form one consistent snapshot. Record and verify the cutoff for every extract and measure.
- Calling a common cutoff “equal maturity.” A common valuation date gives older periods more elapsed time than recent ones. Calculate period age explicitly.
- Treating an age threshold as proof of completeness. A screening rule is only as meaningful as its stated purpose and support. Age alone does not establish that no further claims or changes will occur.
- Inferring an ultimate result from observed data alone. The snapshot shows what is recorded by the cutoff. It does not reveal every claim or final amount that may emerge later.
Check Your Understanding
For each problem, apply the stated date convention and distinguish what is observable from what is mature.
- A claim occurred October 1, 2025 and was reported February 1, 2026. A second claim occurred January 1, 2026 and was reported May 1, 2026. At an April 1, 2026 valuation date, which claims are included, and what is the first claim's age in whole months? (Answer: Only the first claim is included; it is 6 months old.)
- A December 31, 2025 valuation extract contains a claim reported December 30, 2025 and another reported January 2, 2026. How many are observable at that cutoff? (Answer: One claim.)
- At September 30, 2026, two occurrence periods ended March 31, 2026 and June 30, 2026. Give their ages in whole months and identify which is more mature. (Answer: 6 months and 3 months, respectively; the period ending March 31 is more mature.)
- A comparison combines one period's paid amount valued at December 31, 2026 with another period's paid amount valued at June 30, 2026. What is the date problem, and what should the analyst do? (Answer: The amounts mix valuation dates; select or rebuild both paid values at one common valuation date.)
- A recent quarter has 100 policy-years and 3 claims reported by the valuation date; an older quarter has 100 policy-years and 5 reported claims. Can the observed rates alone establish that the recent quarter has lower underlying frequency? (Answer: No; the recent quarter may be less mature, so its observed claim count may be incomplete.)