Why One Set of Claims Can Produce Different Annual Results
A claim has more than one relevant date. The loss occurs on one date, and the insurer may receive its first report of the claim later. If results are grouped by occurrence, the claim belongs to the accident year in which the loss happened. If results are grouped by first report, it belongs to the report year in which the insurer first received notice. Reporting delays can therefore shift a claim from one annual total to another, even though the claim itself has not changed.
This distinction matters whenever a reader compares period results. A higher report-year total does not necessarily mean that more losses occurred in that year: it may include delayed reports from earlier accidents. Conversely, an accident-year total can include claims that were not reported until after the year of occurrence. To make a meaningful comparison, first identify the date used to assign claims to periods and the claim amount being summarized.
Define the Claim Measure Before Grouping
A period total is meaningful only after specifying what is being totaled. It might be claim counts, reported amounts, or another consistently defined claim measure. For the examples below, each claim has one fixed amount, and the same amount is used in both classifications. The amounts are hypothetical and are treated as measured on the same basis, so differences between the two tables arise only from the choice of year.
For a set of claims indexed by \(j\), let \(a_j\) be the accident year, \(r_j\) the first-report year, and \(x_j\) the claim amount being summarized. The accident-year amount for year \(y\) is the sum of amounts for claims with \(a_j=y\). The report-year amount for year \(y\) is the sum for claims with \(r_j=y\).
Here \(\mathbf{1}(\text{condition})\) equals 1 when the condition is true and 0 otherwise. For counts, replace each \(x_j\) with 1. The labels “accident year” and “report year” identify the grouping rule; they do not, by themselves, specify whether \(x_j\) is a count, a reported estimate, or some other claim measure.
A report-year grouping is also not the same as a calendar-year paid-loss grouping. In the latter, a payment is assigned to the year in which cash was paid. In a report-year grouping, a claim is assigned once, according to its first-report date; later payments do not create a new first report. The measure and assignment rule must both be stated.
Worked Example: Reclassifying Six Claims by Year
A hypothetical insurer reviews six claims. Each amount is a fixed value in thousands of dollars on a common basis. Reclassify the claims by accident year and by first-report year, and check that the grand totals agree.
| Claim | Accident year | First-report year | Amount ($000) |
|---|---|---|---|
| A | 2023 | 2023 | 12 |
| B | 2023 | 2024 | 8 |
| C | 2024 | 2024 | 15 |
| D | 2024 | 2025 | 10 |
| E | 2025 | 2025 | 6 |
| F | 2023 | 2025 | 4 |
Method. Sum each row amount under the year named by the chosen date. No interest, mortality, or probability model is used; this is a deterministic reclassification of a hypothetical claim list.
Accident-year totals. For 2023, claims A, B, and F give \(12+8+4=24\) thousand dollars. For 2024, claims C and D give \(15+10=25\) thousand dollars. For 2025, claim E gives 6 thousand dollars.
Report-year totals. For 2023, claim A gives 12 thousand dollars. For 2024, claims B and C give \(8+15=23\) thousand dollars. For 2025, claims D, E, and F give \(10+6+4=20\) thousand dollars.
Check. The accident-year grand total is \(24+25+6=55\) thousand dollars. The report-year grand total is \(12+23+20=55\) thousand dollars. The group totals differ by year, but the total over all claims is the same.
Common mistake avoided: Claim F is reported in 2025 but arose from a 2023 accident. Putting its 4 thousand dollars in 2025 in both tables would incorrectly use report year in the accident-year classification.
Why the Grand Totals Reconcile
If the same complete set of claims is included in both views, and each claim has one accident year and one first-report year, reclassification cannot change the total amount. It changes only which group receives each claim. To see why, sum the accident-year totals over all years. Each claim contributes \(x_j\) once, because it has exactly one accident year. The same argument applies to report years.
The middle equality holds because, for each claim \(j\), exactly one year \(y\) satisfies \(a_j=y\), so the sum of its accident-year indicators is 1. The corresponding sum of report-year indicators is also 1. This reconciliation requires the same claim population, the same amount definition, and a period range that includes every relevant claim. If one view omits late reports, or if the amounts are measured on different bases, the totals need not match.
Reporting Delays Shift Claims Across Periods
The reporting lag is the time between the loss occurrence and the insurer’s first receipt of notice. A claim reported in its accident year has no year-boundary shift in annual grouping. A claim reported in a later year moves from its accident-year total into that later report-year total. In a multi-year portfolio, a report-year total can therefore combine current-year accidents with delayed reports from earlier accident years.
Worked Example: Measuring the Effect of a One-Year Delay
A hypothetical portfolio has the following claim counts and amounts, grouped by accident year and then split according to first-report year. The amounts are fixed thousands of dollars; all claims are included, and no interest or probability model is assumed. Find each year’s totals under both groupings and identify the contribution from delayed reports.
| Accident year | First-report year | Claim count | Amount ($000) |
|---|---|---|---|
| 2023 | 2023 | 2 | 30 |
| 2023 | 2024 | 1 | 20 |
| 2024 | 2024 | 2 | 50 |
| 2024 | 2025 | 1 | 40 |
| 2025 | 2025 | 1 | 10 |
Accident-year calculation. In 2023, the count is \(2+1=3\) and the amount is \(30+20=50\) thousand dollars. In 2024, the count is \(2+1=3\) and the amount is \(50+40=90\) thousand dollars. In 2025, the count is 1 and the amount is 10 thousand dollars.
Report-year calculation. In 2023, there are 2 claims for 30 thousand dollars. In 2024, the total includes the one delayed 2023 claim and the two claims from 2024: the count is \(1+2=3\), and the amount is \(20+50=70\) thousand dollars. In 2025, the total includes the delayed 2024 claim and the 2025 claim: the count is \(1+1=2\), and the amount is \(40+10=50\) thousand dollars.
Check. Accident-year counts total \(3+3+1=7\), and report-year counts total \(2+3+2=7\). Amounts total \(50+90+10=150\) thousand dollars under accident year and \(30+70+50=150\) thousand dollars under report year. The report-year amount of 70 in 2024 is not all attributable to 2024 accidents: 20 comes from a 2023 accident.
Common mistake avoided: Treating every claim reported during 2024 as a 2024 accident would incorrectly move the 20 thousand dollar delayed claim into accident year 2024.
A Period-Allocation View
A useful way to organize the relationship is to make a table whose rows are accident years and whose columns are report years. Each cell contains the claims from that accident year first reported in that report year. Summing across a row gives the accident-year total; summing down a column gives the report-year total. This is a classification table, not a development method: it describes where the observed claims are assigned under the two date rules.
Worked Example: Building Report-Year Totals From Allocation Percentages
For a hypothetical portfolio, suppose the amounts in each accident year are allocated across report years as follows. These are assumed proportions of amount for this example; they are not asserted as empirical reporting patterns. Each row’s percentages sum to 100%, and the amounts are all measured on the same fixed basis. Calculate the report-year totals.
| Accident year | Total amount ($000) | Reported same year | Reported one year later |
|---|---|---|---|
| 2022 | 100 | 70% | 30% |
| 2023 | 120 | 80% | 20% |
| 2024 | 150 | 80% | 20% |
Method. Multiply each accident-year amount by the proportion assigned to each report year. Because only same-year and one-year-later reports are specified, no amount is assigned to any other year.
Report year 2022. The 2022 accident-year amount reported in 2022 is \(100(0.70)=70\) thousand dollars. Thus the report-year 2022 total is 70 thousand dollars.
Report year 2023. The delayed portion from accident year 2022 is \(100(0.30)=30\). The same-year portion from accident year 2023 is \(120(0.80)=96\). Therefore, report-year 2023 totals \(30+96=126\) thousand dollars.
Report year 2024. The delayed portion from accident year 2023 is \(120(0.20)=24\). The same-year portion from accident year 2024 is \(150(0.80)=120\). Therefore, report-year 2024 totals \(24+120=144\) thousand dollars.
Report year 2025. The delayed portion from accident year 2024 is \(150(0.20)=30\) thousand dollars.
Check. Accident-year amounts total \(100+120+150=370\) thousand dollars. Report-year amounts total \(70+126+144+30=370\) thousand dollars. The 2025 report-year amount is present even though there is no 2025 accident-year amount in the data, because 30 thousand dollars from 2024 is reported one year later.
Common mistake avoided: Applying each percentage to the report year’s total rather than to its accident-year amount would reverse the allocation direction. The percentages in each row describe how that accident year’s amount is distributed across report years.
Interpretation and Common Pitfalls
Neither grouping is automatically the correct answer for every question. Accident-year grouping connects claims to the year of the loss event. Report-year grouping connects them to when the insurer first learned of the claims. The appropriate view depends on what period question is being asked. In either case, state the claim measure, the assignment date, and the observation window.
- Confusing first report with payment. A claim first reported in 2023 and paid in 2024 remains assigned to report year 2023 under a first-report grouping.
- Confusing claim occurrence with notice. A 2023 accident first reported in 2024 belongs to accident year 2023 and report year 2024.
- Assuming annual totals must match. The two totals for a particular year can differ because the included claims differ. It is the all-period total that reconciles, subject to a common complete claim set and consistent amounts.
- Ignoring the observation boundary. Claims from the final accident year may be reported after the data cutoff. A report-year table can show those later reports, while an incomplete dataset may not yet contain them.
- Changing the amount basis unnoticed. If claim amounts are revised between measurements, differences may reflect both classification and measurement changes. The simple reconciliation shown here assumes the same fixed amount per claim in both views.
Check Your Understanding
For each problem, classify the claims using the stated date rule and check the arithmetic before comparing totals.
- A hypothetical claim has an accident date of December 28, 2024, and is first reported on January 3, 2025. State its accident year and report year. (Answer: Accident year 2024; report year 2025.)
- Three claims have these accident year, first-report year, and amount values in thousands: (2022, 2022, 5), (2022, 2023, 7), and (2023, 2023, 9). Find the 2023 accident-year amount and report-year amount. (Answer: Accident year 2023 amount = 9 thousand dollars; report year 2023 amount = 16 thousand dollars.)
- A hypothetical set has 4 claims totaling 40 thousand dollars from accidents in 2024, of which 3 claims totaling 28 thousand dollars are first reported in 2024; the remaining claim is first reported in 2025. Find its accident-year and report-year count and amount contributions. (Answer: Accident year 2024: 4 claims, 40 thousand dollars; report year 2024: 3 claims, 28 thousand dollars; report year 2025: 1 claim, 12 thousand dollars.)
- Accident-year amounts are 80 thousand dollars for 2023 and 100 thousand dollars for 2024. Assume 25% of the 2023 amount is reported in 2024, while 90% of the 2024 amount is reported in 2024 and 10% in 2025. Find report-year totals for 2023, 2024, and 2025. (Answer: 2023: 60 thousand dollars; 2024: \(20+90=110\) thousand dollars; 2025: 10 thousand dollars.)
- A complete claim list has amounts totaling 240 thousand dollars. The accident-year table totals 155 thousand dollars for years 2022–2024 and 85 thousand dollars for 2025. The report-year table totals 70 thousand dollars for years 2022–2024. What must the 2025 report-year total be if both tables use the same complete claims and fixed amounts? (Answer: \(240-70=170\) thousand dollars.)