A hospital CFO reviewing the professional liability program sees a claims-made policy, renewed every year, and a loss run that shows every open claim with a reserve against it. Then the auditor asks where the tail liability is, and the actuary’s report carries a line called “unreported claims” that no loss run supports. The number comes from how a claims-made policy is built.
This explainer covers what the claims-made trigger does, who owns the claims that have happened but not yet surfaced, how that liability is estimated, how FASB guidance treats it, and what changes when the claims-made policy is written by your own captive.
One vocabulary warning first. “Tail” has two unrelated meanings in reserve work. A development tail factor is the multiplier that carries losses from the last column of a triangle to ultimate; that is the subject of tail factor selection for captives. This article is about the other tail: the reporting tail of a claims-made program, and the coverage and the liability named after it.
Two triggers: occurrence and claims-made
An occurrence policy covers events that happen during the policy period, no matter when the claim is reported. A 2020 occurrence policy still responds to a claim first reported in 2027, as long as the event took place in 2020.
A claims-made policy covers claims reported during the policy period. The CAS ratemaking text by Werner and Modlin puts the difference in one line: the coverage trigger is the date the claim is reported rather than the date the event occurs. A 2020 claims-made policy that expired without renewal does nothing for a claim reported in 2027, even though the event happened while the policy was in force.
Claims-made was the insurance industry’s answer to a pricing problem, not a drafting preference. Marker and Mohl, writing for the Casualty Actuarial Society in 1980 about the St. Paul’s medical liability program, describe a line that had grown large enough to threaten the company “if we guessed wrong about the IBNR,” and a decision to find a way to take on that risk a year at a time and price it a year at a time.
The fourth of Marker and Mohl’s five principles of claims-made ratemaking is the one that matters for your balance sheet: claims-made policies incur no liability for IBNR. Werner and Modlin restate it more precisely. A claims-made policy has no pure IBNR component, meaning no claims incurred but not reported. The insurer still has to reserve for development on the claims that were reported, the piece often called IBNER.
The risk of late-reported claims did not vanish when the industry changed forms. It moved to the policyholder.
Retroactive dates and extended reporting periods
Two policy features decide how far a claims-made policy reaches.
The retroactive date. A claims-made policy covers only claims arising from incidents on or after its retroactive date. When an organization first moves from occurrence to claims-made coverage, the retroactive date should be coordinated with the expiration of the last occurrence policy so there is neither a gap nor an overlap. At each renewal the retroactive date should stay where it was. If a new insurer resets it to the new policy’s inception date, every unreported incident from the earlier years loses its coverage.
The extended reporting period, or tail coverage. When the last claims-made policy expires, because the insured switches back to occurrence, changes carriers without keeping its retroactive date, closes, or a physician retires, there is no policy for claims that occurred during the claims-made years but are reported afterward. An extended reporting endorsement, usually called tail coverage, covers exactly those claims. Marker and Mohl note that the St. Paul offered a single-payment reporting endorsement to insureds ending coverage through death, disability, or retirement.
Marker and Mohl also describe what tail coverage is in actuarial terms: the insurer is “essentially selling IBNR coverage,” and pricing a reporting endorsement “is equivalent to the determination of IBNR reserves.” Hold on to that sentence. If you do not buy the tail, the amount the insurer would have reserved is the amount you owe.
One caution from Friedland’s CAS reserving text: policies do not all define the trigger the same way. For some claims-made policies the notification date, not the claim report date, triggers coverage. The policy wording decides which incidents count as reported, and so which ones sit inside the policy and which sit in your tail.
First-year, mature, and step factors
A claims-made policy in its first year, with a retroactive date equal to its inception date, covers very little: only incidents that both happen and get reported within that year. The second-year policy covers incidents from two years, the third from three, and so on. A mature claims-made policy is one whose retroactive date is far enough back that every reporting lag is covered.
Insurers price this progression with step factors, each a percentage of the mature claims-made rate. Werner and Modlin show an illustrative structure, which they describe as a potential step factor structure, not a benchmark:
| Claims-made year | Step factor |
|---|---|
| First | 40% |
| Second | 70% |
| Third | 85% |
| Fourth | 95% |
| Fifth or more | 100% |
Read as a reporting pattern, the table says 40 percent of a mature policy’s cost comes from incidents that occurred and were reported in the same year, 30 percent from incidents one year old, and so on. Real patterns are specific to the program.
The first-year discount is not a saving. It is the cost of the incidents that occurred that year and will be reported later, and that cost is still owed by someone.
Why the unreported claims are your liability
Follow one incident. A patient is injured in March 2026. The claim is first asserted in 2028. If the hospital has renewed its claims-made policy every year with the original retroactive date, the 2028 policy responds, subject to the 2028 limits, deductible, and terms. If the hospital has bought tail coverage, the tail responds. If neither is true, nothing responds.
At December 31, 2026, the incident has happened. The hospital’s obligation to the patient exists whether or not a claim has been made and whether or not any policy will eventually pay. The expired 2026 policy is not on the hook for it. A future policy that has not been purchased yet is not an asset. That is why the unreported claims for incidents through the balance sheet date are the insured’s liability, generally called the tail liability or unreported claims liability.
For a hospital, a physician group, or a captive parent, it has to be estimated separately, because no loss run will ever show it. The hospital professional liability explainer covers the wider reserve picture for health systems; the tail liability is the part of it that claims-made buyers most often miss.
Report-year triangles, and what they cannot see
Most reserve reports organize losses by accident year: all claims from incidents in 2022, tracked as they are reported and paid. Claims-made experience is usually organized by report year instead: all claims reported in 2022, whenever the incident happened. Werner and Modlin lay the two out on one grid of report year against reporting lag. A claims-made policy is a row of that grid. An occurrence policy is a diagonal.
Friedland gives the main advantage of report-year data. The number of claims is fixed at the close of the year, apart from claims reported but not yet recorded, so the data are generally more stable and the development patterns more readily determinable than on an accident-year basis, where the claim count keeps changing. A report-year triangle tracks only how the known claims mature.
Friedland is equally direct about the limitation. Techniques based on report-year data “only measure development on known claims and not pure IBNR,” pure IBNR is frequently the harder part of the estimate, and other methods are required to derive it. So a report-year chain ladder on your claims-made data answers one question: are the claims we know about reserved adequately? It cannot answer the second: how many claims are out there that we do not know about yet?
How the tail liability is estimated
The actuary needs the missing diagonal of the grid: losses from incidents on or before the valuation date that will be reported after it. Marker and Mohl define pure IBNR in exactly those terms, as the sum of losses with a report year after the valuation date and an accident year on or before it. In practice the estimate is usually built one of three ways.
- Occurrence-basis projection. Re-sort the same claims by accident year, project ultimate losses on that basis, and subtract the ultimate value of claims already reported. What remains is the unreported piece.
- Reporting pattern applied to exposure. Estimate the share of each occurrence year’s losses still unreported from the historical lag between incident and report, and apply it to an expected loss per unit of exposure, such as occupied beds or physician full-time equivalents.
- Frequency and severity. Estimate the number of unreported claims from the reporting pattern and multiply by an average cost per claim, trended to the period in which those claims are expected to be reported and settled.
All three depend on an incident-to-report lag pattern, so ask which pattern was used and where it came from. The CAS Statement of Principles on loss reserves, reproduced as an appendix to Friedland’s text, also notes that claims-made data should be kept separate from occurrence-basis experience, and that the contract wording has to be reviewed to decide what reserve, if any, is needed for claims reported after policy expiration.
A worked illustration
The numbers below are hypothetical and chosen to make the arithmetic visible. A hospital starts a claims-made program on January 1, 2022, with that date as its retroactive date. Assume:
- Incidents in each year will ultimately cost $5.0 million, with no growth and no inflation.
- Claims are reported over five years: 40 percent in the year of the incident, then 30, 15, 10, and 5 percent, the same pattern as the textbook step factors above.
What the claims-made policy covers each year is the reported claims. What it leaves behind is the unreported share of every incident year so far.
| Year | Claims-made losses (reported in year) | Tail liability at year end | Change in tail liability | Claims-made plus change |
|---|---|---|---|---|
| 2022 | $2.00M | $3.00M | $3.00M | $5.00M |
| 2023 | $3.50M | $4.50M | $1.50M | $5.00M |
| 2024 | $4.25M | $5.25M | $0.75M | $5.00M |
| 2025 | $4.75M | $5.50M | $0.25M | $5.00M |
| 2026 | $5.00M | $5.50M | $0.00M | $5.00M |
Take the December 31, 2026 tail liability apart: 60 percent of the 2026 incidents are unreported ($3.00 million), 30 percent of 2025 ($1.50 million), 15 percent of 2024 ($0.75 million), and 5 percent of 2023 ($0.25 million). Everything from 2022 has been reported. The total is $5.5 million, or 1.1 years of losses.
Three things to take from the table.
- Claims-made cost plus the change in the tail equals occurrence cost, every year. The hospital that paid for $2.0 million of claims-made losses in 2022 also took on $3.0 million of unfunded tail. The cheap early years were a deferral.
- Once the program matures, the tail liability stops growing only if nothing else changes. In this flat illustration it levels off at $5.5 million. Add exposure growth or rising claim costs and it keeps climbing, because each new incident year is larger than the one rolling off. Werner and Modlin’s own example of the cost gap, with claim costs rising 5 percent a year, shows an occurrence policy costing $1,105.13 against $1,000.00 for the claims-made policy of the same year.
- The tail liability is real money at exit. If this hospital closed or was sold on December 31, 2026, with no tail policy, $5.5 million of claims would arrive over the next four years with no insurer behind them.
A slower reporting pattern makes all of this larger. If only 20 percent of losses were reported in each of five years, the tail liability would be 2.0 years of losses, not 1.1.
The accounting, as far as the FASB documents go
Three points come directly from FASB material. Your auditor will work from the current Codification text, which this article does not quote directly, so treat the following as a map and not as a substitute.
The insured accrues the unreported claims. EITF Issue 03-8, “Accounting for Claims-Made Insurance and Retroactive Insurance Contracts by the Insured Entity,” records the consensus that insured entities, including those that use a claims-made approach, recognize a liability for the probable losses from IBNR claims and incidents when the loss is both probable and reasonably estimable. That is the ASC 450 test applied to the tail. For health care entities, ASC 954-450-25-2 as amended says the ultimate costs of malpractice claims, including the costs of litigating or settling them, are accrued when the incidents that give rise to the claims occur, and it points to Section 720-20-25 and Subtopic 944-40 for the accounting for insurance claims costs including incurred-but-not-reported claims.
A tail premium quote is not the measure of the liability. The same EITF consensus says the estimated cost of purchasing tail coverage is not relevant in determining the loss to be accrued. There is one carve-out: if the insured holds a unilateral option to buy tail coverage at a premium that cannot exceed a specified fixed maximum, it could record a receivable for the portion of the IBNR liability insurable under the tail, after deductibles and limits, and would then record the expected tail premium as a cost. The Task Force added that buying tail coverage does not eliminate the need to consider an additional liability because of policy limits or other factors.
Liabilities and recoveries are shown gross. ASU 2010-24, issued in August 2010, amended Topic 954 so that a health care entity does not net insurance recoveries against the related claim liability and determines the liability without considering recoveries. An indemnified entity recognizes an insurance receivable at the same time, measured on the same basis as the liability, subject to a valuation allowance for uncollectible amounts. The amendments took effect for fiscal years beginning after December 15, 2010. The Task Force’s reasoning was that the health care entity generally remains liable for the claim and is exposed to the insurer’s credit risk. The update also removed older industry language under which the IBNR accrual could effectively be limited to the maximum tail premium.
The EITF abstract also describes the expense pattern for a buyer of prospective claims-made policies: estimated annual expense is the premium, plus the increase in the IBNR liability, less the increase in the related insurance recoverable. That mirrors the last column of the table above.
When the claims-made policy is written by your captive
A captive can issue its policy to the parent on a claims-made form, for example to match the form of the excess coverage above it. That choice splits one economic liability across two balance sheets.
- The captive holds case reserves and development on reported claims, estimated from report-year data. On a pure claims-made policy it has no pure IBNR, which is why a captive-only reserve report can show a surprisingly small IBNR for a long-tailed line.
- The parent still has occurrence-based exposure. Patients were treated, incidents occurred, and the parent owes the unreported claims whether or not the captive’s next policy has been issued. The parent accrues the tail liability.
Three failure modes follow, and each is a scope problem before it is an actuarial one.
- The tail falls between two reports. The actuary is engaged by the captive, opines on the captive’s reserves, and nobody estimates the parent’s tail. The consolidated balance sheet is short by the whole unreported block.
- The tail is counted twice. The captive’s actuary projects ultimates on an accident-year basis out of habit while the parent separately accrues a tail. Check that the two estimates add up to one occurrence-basis number.
- The captive’s terms change. If the captive issues tail coverage, moves to an occurrence form, or resets the retroactive date, liability moves between the entities. The same is true in a wind-down: a loss portfolio transfer or runoff plan that covers only the captive’s reported claims leaves the parent’s tail untouched.
The single-parent captive IBNR guide covers the captive side in more depth. The practical rule is to ask for one exhibit that reconciles captive reserves plus parent tail to a single occurrence-basis estimate.
On the insurer side, the Wire’s piece on hospital earnings and malpractice severity discusses adverse development on commercial claims-made business, a reminder that reported claims still develop.
Questions to ask your actuary
- Is the estimate on a claims-made or an occurrence basis, and which entity is each piece for: the captive, the parent, or both?
- What is the retroactive date on every policy in the tower, and has it ever moved?
- How does each policy define a reported claim, and do our incident reports to the insurer or captive match that definition?
- What incident-to-report lag pattern supports the tail liability, is it ours or a benchmark, and what happens to the number if reporting is one year slower?
- Does the tail liability include defense costs, and is it gross of any deductible, self-insured retention, and policy limits?
- Do we hold a contractual right to buy tail coverage at a capped premium, and how was that treated in the estimate and the accrual?
- If the program ended at the valuation date, what would the unreported block be, and how does that compare with a tail quote?
Sources
- Basic Ratemaking, Fifth Edition, Chapter 16: Claims-Made Ratemaking, Geoff Werner and Claudine Modlin, Casualty Actuarial Society, May 2016
- Rating Claims-Made Insurance Policies, Joseph O. Marker and F. James Mohl, Casualty Actuarial Society Discussion Paper Program, 1980
- Estimating Unpaid Claims Using Basic Techniques, Version 3, Jacqueline Friedland, Casualty Actuarial Society, July 30, 2010
- EITF Abstracts Issue No. 03-8, Accounting for Claims-Made Insurance and Retroactive Insurance Contracts by the Insured Entity, Financial Accounting Standards Board
- Accounting Standards Update No. 2010-24, Health Care Entities (Topic 954): Presentation of Insurance Claims and Related Insurance Recoveries, Financial Accounting Standards Board, August 2010
Frequently asked questions
What is tail coverage on a claims-made policy?
Tail coverage, also called an extended reporting endorsement, covers claims that occurred while claims-made coverage was in force but are reported after the last claims-made policy expires. Without it, or a renewal that keeps the same retroactive date, those late-reported claims have no policy to respond to them.
Does a claims-made policy cover IBNR?
Not pure IBNR. A claims-made policy covers claims reported during its term, so it carries development on those reported claims but no liability for claims reported after it ends. The unreported claims for incidents that have already happened remain the insured's exposure until a renewal or tail policy picks them up.
Do you have to accrue a liability for unreported claims if you have claims-made insurance?
Yes, when the loss is probable and reasonably estimable. FASB's EITF Issue 03-8 reached that consensus for insured entities using claims-made coverage, and ASC 954-450-25-2 tells health care entities to accrue the ultimate costs of malpractice claims when the incidents that give rise to the claims occur.
Why is claims-made data organized by report year instead of accident year?
Because the report date is what triggers coverage. A report year groups every claim reported in that year, the count is fixed once the year closes, and development patterns are generally more stable than on an accident-year basis. The trade-off is that report-year analysis says nothing about claims not yet reported.