Why Claims Processes Matter to Loss Development
A loss triangle records amounts at different development ages, but those amounts are produced by business processes as well as by the underlying claims. A change in payment speed can alter paid development without changing the total cost of claims. A change in case-reserving guidance can alter incurred development. A revised data extract can change which payments, recoveries, or expenses are included. If periods on either side of such a change are treated as comparable without investigation, selected development patterns may reflect the process change rather than the claims.
The task is not to assume that every operational change invalidates a triangle. It is to identify a plausible break, determine which data and periods it affects, and establish whether the amounts can be put on a consistent basis. A change in operations may have no material effect, while a subtle definition change may affect every observation in a column. The evidence and the data definition matter more than the label given to the change.
How a Process Change Enters a Development Factor
Let \(C_j\) denote cumulative paid claims at development age \(j\), and let \(U\) denote the eventual paid amount for a fixed group of claims. In a simplified illustration, suppose the expected cumulative paid amount at age \(j\) is \(q_jU\), where \(q_j\) is the proportion of the eventual amount paid by that age. If the process and data basis are consistent across ages, the adjacent-age ratio is approximately:
The eventual amount cancels, so the ratio describes relative emergence between ages in this simplified setting. But a change in payment procedures can change the proportions \(q_j\). For example, moving routine payments earlier may increase the proportion paid at the earlier age. The observed ratio can then fall even if the eventual cost \(U\) is unchanged. This is why a changed factor is a diagnostic signal, not by itself proof that claim severity has changed.
The same reasoning applies to incurred amounts, but the recorded amount also depends on case reserves. Incurred claims are generally made up of paid claims plus case reserves on the data basis being used. A new reserving guideline can therefore shift incurred development even if paid amounts and eventual claim costs are stable. Before interpreting a change in a factor, ask what process could have changed each component of the numerator and denominator.
Identify when the new procedure, system, or definition began, and which accident periods, development ages, or claim segments could be affected.
Record whether the change concerns payment timing, case-reserve adequacy, claim closure, recoveries, expenses, or the way amounts are extracted and grouped.
Compare relevant paid and incurred observations, claim counts or operational summaries when available, and the documentation defining the amounts.
Use consistently restated data if available; otherwise consider separating affected periods or treating selected observations cautiously. Document the decision and its limitations.
Diagnosing Handling and Reserving Changes
Claims-handling changes can affect when amounts appear in a triangle. Examples include a new settlement authority that speeds payment, a backlog that delays processing, a change in the use of partial payments, or a new closure-and-reopening procedure. These changes may influence paid amounts, incurred amounts, or both. The direction is not guaranteed: faster processing might increase early paid emergence, but a change in settlement mix could also change which claims are paid first.
Case-reserving changes have a more direct effect on incurred amounts. For example, revised guidance might require higher initial reserves, more frequent reserve reviews, or a different approach to claims with uncertain outcomes. In that situation, early incurred values may increase even if paid values do not. Later case-reserve releases may then create a different incurred pattern. The analyst should not interpret the resulting incurred factors as if they were independent of the reserve-setting process.
A useful diagnosis compares the affected measure with evidence that should respond differently to the change. If a reported payment-processing change coincides with faster early paid development but no comparable movement in the incurred pattern, that is consistent with a timing effect, though it does not prove one. If paid values appear stable while incurred values shift at the date new reserving guidance began, a case-reserving explanation deserves investigation. Neither comparison substitutes for verifying the claim data and operational history.
Worked Example: A New Payment Workflow
A hypothetical insurer changes its workflow for small commercial-property claims. For one mature group handled under the old workflow, cumulative paid amounts are $80,000 at 12 months and $136,000 at 24 months. For a comparable group handled under the new workflow, the amounts are $100,000 and $145,000. Calculate each 12-to-24-month ratio, then assess what the difference can and cannot establish.
Method. Calculate the adjacent-age ratio separately for each workflow, using the same development ages and the same paid-amount basis.
Old workflow. The 12-to-24-month ratio is:
The paid amount grew by $56,000 over the interval, since \(\$136{,}000-\$80{,}000=\$56{,}000\). Relative to the 12-month amount, that growth is \(\$56{,}000/\$80{,}000=0.70\), giving a ratio of \(1+0.70=1.70\).
New workflow. The corresponding ratio is:
Here the increase is \(\$145{,}000-\$100{,}000=\$45{,}000\), and \(\$45{,}000/\$100{,}000=0.45\), giving \(1+0.45=1.45\). The new-workflow ratio is lower by \(1.70-1.45=0.25\).
This difference is consistent with the possibility that more payments occur by 12 months under the new workflow, leaving less relative emergence between 12 and 24 months. It does not establish that the new group has lower ultimate costs: the two groups have different 12-month amounts, and the example supplies no evidence about their eventual totals or claim mix. The next check would be to verify that the groups are comparable and that the payment definition is unchanged.
Check and common mistake. The ratios use the 12-month amount as the denominator in both cases and reproduce the stated 24-month amounts when multiplied back: \(\$80{,}000(1.70)=\$136{,}000\), and \(\$100{,}000(1.45)=\$145{,}000\). A common mistake is to call the lower ratio proof of improved claim outcomes. It measures a change in relative paid emergence, not a demonstrated change in eventual cost.
Worked Example: A Change in Case-Reserve Guidance
A hypothetical liability insurer reviews its case-reserving instructions. For a group using the former instructions, at 12 months paid claims are $75,000 and case reserves are $45,000; at 24 months paid claims are $110,000 and case reserves are $25,000. For a group using the revised instructions, the 12-month paid amount is $75,000 and the case reserve is $75,000. Assume incurred amount equals paid amount plus case reserves, with all amounts on the same basis. Calculate the former 12-to-24-month incurred ratio and the projection obtained by applying it to the revised group’s 12-month incurred amount. Then state the limitation of that calculation.
Method. First calculate incurred amounts from paid claims plus case reserves. Then divide the former group’s 24-month incurred amount by its 12-month incurred amount, and apply that ratio to the revised group’s 12-month incurred amount.
For the former instructions, 12-month incurred is:
At 24 months, incurred is:
The former-instructions incurred ratio is therefore:
Under the revised instructions, 12-month incurred is \(\$75{,}000+\$75{,}000=\$150{,}000\). Applying the former ratio gives:
This is a mechanical projection using the former ratio; it is not evidence that $168,750 is a reliable estimate for the revised group. The higher revised case reserve may reflect the new instructions rather than a higher eventual claim cost. The analyst would need evidence about reserve adequacy and subsequent development before concluding that the former factor transfers.
Check and common mistake. The projected amount is 12.5% above $150,000, because \(0.125(\$150{,}000)=\$18{,}750\), and \(\$150{,}000+\$18{,}750=\$168{,}750\). A common mistake is to compare the revised 12-month incurred amount directly with the former 24-month amount and treat the difference as a reserve shortfall. The groups use different reserving instructions, so that comparison confounds case-reserving practice with development.
Definition Changes: Make the Measurement Basis Explicit
A data definition can change without any operational change in how claims are handled. Examples include switching between gross and net amounts, including or excluding allocated expenses, changing how salvage or subrogation recoveries are recorded, or changing which claim statuses enter the extract. A change in the database field name alone is not enough to establish equivalence. Confirm what each field contains, the effective date of the definition, and whether historical data were restated.
Comparisons must use a consistent basis. A paid amount net of recoveries is not directly comparable to a gross paid amount if the recovery treatment differs. Likewise, a triangle that includes expenses in newer periods but excludes them in older periods does not measure the same quantity. If data can be restated, the restatement should be checked against documented definitions and reconciled totals. If not, the affected observations should be clearly identified rather than silently treated as comparable.
Worked Example: Gross and Net Amounts in Different Ages
A hypothetical claims extract reports 12-month paid amounts net of recoveries but 24-month amounts gross of recoveries. At 12 months, gross payments are $250,000 and recoveries are $25,000. At 24 months, gross payments are $360,000 and recoveries are $30,000. Calculate the correct net 12-to-24-month ratio and the ratio produced by mistakenly mixing a gross numerator with a net denominator.
Method. Put both ages on a net basis by subtracting recoveries from gross payments, then calculate the ratio. For the incorrect comparison, retain the gross 24-month numerator and the net 12-month denominator.
The 12-month net amount is:
The 24-month net amount is:
The consistent net ratio is:
to four decimal places. If the gross 24-month amount is incorrectly divided by the net 12-month amount, the result is:
The mixed-basis ratio is higher by \(1.6000-1.4667=0.1333\), using the displayed rounded ratios. The difference occurs because the numerator includes $30,000 of recoveries that the net denominator excludes.
Check and common mistake. The net amounts retain the recovery treatment at both ages, and \(330{,}000/225{,}000=22/15=1.4666\ldots\), which rounds to 1.4667. A common mistake is to use the gross 24-month value merely because it is the latest available field. A newer observation is not comparable unless it has the same definition or is converted to a consistent basis.
Choosing a Response and Avoiding Overcorrection
After identifying a potential break, select a response supported by the data. If the source system can provide a consistent historical extract, use and validate that extract. If only certain segments were affected, separating those segments may be more defensible than changing every observation. If there is insufficient evidence to quantify an adjustment, document the break and explain how it limits the interpretation of the affected data. An operational explanation alone does not supply a numerical correction.
Before applying any adjustment, determine whether the change affects levels, timing, definitions, or the population of claims. A correction designed for one effect may worsen another. For instance, adjusting early paid amounts for faster payment handling does not correct a change from gross to net reporting. Any adjustment should be transparent, reproducible, and checked against the original data and the stated measurement basis.
| Observed signal | Possible explanation to investigate | Evidence to seek |
|---|---|---|
| Early paid amounts rise relative to later development | Faster payment processing or a change in payment practices | Workflow dates, payment timing, and consistent paid definitions |
| Incurred amounts shift while paid amounts appear stable | Changed case-reserving instructions or reserve review practices | Reserve guidance, case-reserve components, and subsequent revisions |
| Amounts change at a system or extract transition | Changed inclusion rules, status filters, or data definitions | Field specifications, extract logic, and reconciliations across systems |
| Development differs for only one claim group | A segment-specific operational change or a change in claim mix | Segment definitions and group-level operational records |
A common exam trap is to see an unusual factor and immediately alter it, or to attribute it to an operational change solely because one occurred nearby in time. The factor is a symptom to investigate. Establish the affected periods and measure, test the proposed explanation against the data, and distinguish evidence of a process change from evidence of a change in underlying claim costs. This diagnostic discipline is aligned with the SOA Exam ASTAM (Advanced Short-Term Actuarial Mathematics) syllabus.
Check Your Understanding
For each hypothetical situation, identify the calculation or diagnostic conclusion that is justified by the stated information.
- Under an old payment process, a group has cumulative paid amounts of $80,000 at 12 months and $136,000 at 24 months. Under a new process, the corresponding amounts are $100,000 and $145,000. Calculate both ratios and their difference. (Answer: Old ratio \(=1.70\); new ratio \(=1.45\); difference \(=0.25\).)
- At 12 months, a group has paid claims of $75,000 and case reserves of $75,000. The former incurred ratio is 1.125. Calculate the mechanical projection from 12 to 24 months, and state what the calculation does not establish. (Answer: 12-month incurred \(=\$150{,}000\); projection \(=\$168{,}750\); it does not establish that the former ratio is appropriate under revised reserving guidance.)
- An extract gives 12-month gross payments of $250,000 and recoveries of $25,000, and 24-month gross payments of $360,000 and recoveries of $30,000. Calculate the ratio on a consistent net basis and the incorrect mixed-basis ratio. (Answer: Net ratio \(=330{,}000/225{,}000=1.4667\); mixed-basis ratio \(=360{,}000/225{,}000=1.6000\).)
- A new reserve review procedure began at the same time incurred amounts increased, while paid amounts showed no similar movement. What is a reasonable diagnostic conclusion, and what should be checked before changing selected development? (Answer: The pattern is consistent with a case-reserving effect, but is not proof; check the reserve guidance, paid and case-reserve components, and whether the data definitions and affected periods are consistent.)
- A payment system change affects only large claims, but a triangle combines large and small claims. What should be established before applying an adjustment to the combined triangle? (Answer: Establish which observations and periods are affected and whether the large-claim segment can be isolated or consistently restated; an adjustment to the combined triangle is not justified by the change alone.)