Workmen's compensation reserving with late-reported large claims
You are deputy to the appointed actuary of an Indian insurer that began writing workmen's compensation (WC) and public liability in 2021-22. Gross written premium rose from ₹850m to ₹2.6bn in three years. A third-party administrator sets case reserves at a flat 70% of the initial claim estimate for every notification. In the latest quarter, 18 claims above ₹10m were reported, 12 of which occurred more than 18 months earlier. Paid claims to date are ₹410m and case reserves ₹620m. Your paid chain-ladder, using six development periods, indicates an ultimate of ₹2.9bn. The largest claim, a factory fire, carries a ₹180m case reserve that the TPA believes is only 50% adequate. The CFO wants to release ₹250m of IBNR to lift the solvency ratio from 1.62 to 1.70; the internal target is 1.70 and the regulatory minimum is 1.50. The board meets in ten days.
1. Estimate a reasonable range for the ultimate claims and explain the method you would use, including the treatment of the large claim.
Read the worked answer
Reported claims are ₹410m paid plus ₹620m case reserves = ₹1,030m. The paid chain-ladder gives ₹2.9bn, implying IBNR of ₹1.87bn. That is plausible for a long-tail liability book, but the development pattern has only six periods and is dominated by a few large claims, so the chain-ladder is unstable.
Build a range using at least three methods. First, a Bornhuetter-Ferguson estimate: if earned premium is about ₹2.4bn and the expected loss ratio is 75%, the expected ultimate is ₹1.8bn. If only 35% is reported, BF gives 1,030 + 0.65 × (1,800 − 1,030) = ₹1,530m. Second, a frequency-severity estimate: project reported claim counts to ultimate and apply an average severity loaded for the late-reporting tail. Third, a case-reserve adequacy study: the flat 70% rule is not evidence-based and is likely to understate severe claims.
The factory fire reserve is ₹180m and the TPA believes it is 50% adequate, so add at least ₹90m. Late reporting of 12 large claims suggests IBNR is higher than a paid-only method shows. A reasonable range is ₹2.0–3.0bn, with a point estimate around ₹2.4–2.6bn. Do not use the chain-ladder 2.9bn as a point estimate without testing its sensitivity to the loss-development tail.
2. Evaluate the CFO's proposed ₹250m IBNR release and advise the board, including the professional and regulatory considerations.
Read the worked answer
A release is only justified if the best estimate of ultimate claims has fallen. The evidence points the other way: late-reported large claims, an inadequate flat case-reserve rule, a large claim that is probably under-reserved by at least ₹90m, and a chain-ladder estimate above the expected-loss-ratio estimate. Releasing ₹250m to lift the solvency ratio from 1.62 to 1.70 is reverse-engineering the reserve to a target. That is not a best estimate and would misstate liabilities and capital.
Advise the board that the solvency ratio should be managed through capital, reinsurance or business actions, not through reserve releases. If the board wants a release, require a claim-by-claim review, an independent peer review and a documented change in the expected ultimate. The appointed actuary must be able to sign the reserve opinion without pressure. Under the regulatory framework, the appointed actuary has a statutory responsibility to certify reserves and solvency; a release that flatters the ratio could breach those duties and the profession's code of conduct.
Practical recommendation: present the board with the reserve range, the sensitivity of the solvency ratio to the point estimate, and the actions available if the ratio is below target. If the CFO persists, document the disagreement and escalate to the audit committee.
3. What data and process improvements would you recommend before the next valuation?
Read the worked answer
Improve claim-level data first: notification date, occurrence date, report lag, injury type, industry, wage inflation and recovery/ subrogation. Reconcile paid and case data to the general ledger. Replace the flat 70% case-reserve rule with a claim-specific reserve methodology: initial reserve based on injury severity and expected wage replacement, then reviewed at set milestones. Require TPA authority limits and independent review of claims above a threshold.
Build a reserving triangle by accident quarter and by claim size band, not just in total, so that large-claim development is visible. Study reporting delays and the tail. Run a case-reserve adequacy test comparing initial reserves with ultimate payments. Use BF, frequency-severity and chain-ladder, and document the selection.
Process: establish a reserving committee with claims, actuarial, finance and internal audit; require peer review of the valuation; document data limitations and judgements; and monitor solvency sensitivity. Because the book is growing quickly, expected-loss-ratio methods need a credible benchmark; use industry data and reinsurer pricing where available. Finally, test the reserve range against the solvency ratio so the board sees the connection.