Your reserves developed adversely again. Your actuary’s report has forty pages of triangles and a one-paragraph explanation that does not quite tell you why. This guide is for the risk manager or CFO who wants to know what is actually driving the number and what to do about it.
This is not a reserving primer. If you need the foundational concept first, start with the plain-English IBNR explainer and come back. This article assumes you already know IBNR is a number on your balance sheet that you do not like, and it walks you through how to figure out which part of your claims operation is making it worse.
Adverse reserve development is not random. It has identifiable causes, and those causes have different fixes. The difference between a CFO who accepts adverse development as a cost of doing business and one who actually reduces it over time is almost always diagnostic discipline: knowing which of three things went wrong, and asking the right question about the right one.
The one mental model that matters: case reserves, payment patterns, and mix
Every actuary who has ever shown you a development triangle has made a bet. The bet is that claims recorded to date will continue to develop in a manner consistent with historical patterns (Friedland, p. 84). When that bet is wrong, the triangle produces a bad answer. Adverse development is what happens when reality turns out different from the pattern.
The useful question is not “why was the estimate wrong?” It is “which assumption failed?” Friedland’s study note on basic reserving techniques builds its diagnostic framework around a table of failure modes that maps specific environmental changes to specific method failures (Friedland, Section 4). Every entry in that table falls into one of three buckets, and understanding which bucket your adverse development comes from is the single most important diagnostic step you can take.
Bucket 1: Case reserve adequacy changed. Your claims team started reserving claims differently. Maybe they got more conservative after a regulatory exam. Maybe caseloads grew and adjusters are under-reserving because they do not have time for proper file reviews. Either way, the gap between case reserves and ultimate payments shifted, and the development technique that relies on reported claims is seeing the shift as development instead of recognizing it as a change in practice.
Bucket 2: Payment patterns shifted. Claims are closing faster or slower than historical patterns predicted. A new claims management initiative pushed faster settlements. A TPA change introduced processing delays. A state legislature changed the medical fee schedule or indemnity duration rules. The development technique that relies on paid claims was calibrated to the old speed and is now wrong.
Bucket 3: Claim mix shifted. The types of claims you are generating have changed. A manufacturer added a product line with higher injury severity. A trucking operation expanded into long-haul routes. A hospital opened a higher-acuity unit. Historical development patterns no longer describe the current book because the current book is not the same book.
Each bucket has different symptoms, different data signatures, and different operational fixes. The diagnostics that follow are structured around identifying which one is in play. Often more than one is active, but even knowing the primary driver puts you ahead of most reserve conversations.
Diagnostic 1: Is your case reserve adequacy slipping?
This is the most common driver of adverse development on a reported (incurred) chain ladder estimate, and it is the one most often missed in actuarial review conversations because neither party wants to say the quiet part out loud: someone changed how claims are being reserved, and nobody told the actuary.
Friedland’s failure-mode table flags both case reserve strengthening and case reserve weakening as conditions that break the development technique’s core assumption (Friedland, Section 4). Strengthening means adjusters are posting higher initial reserves, which makes recent accident years look worse than they will ultimately be. Weakening means initial reserves are lower than they should be, which makes recent years look artificially good until late development reveals the gap.
How to spot it. Three signals, all available from data your TPA already produces:
-
Ratio of paid to incurred by maturity age. In a stable environment, the fraction of reported losses that have been paid at any given development age should be roughly consistent across accident years. If that ratio is trending downward for recent years at the same maturity, case reserves are growing relative to payments. Something changed in reserving practice.
-
Average case outstanding by development age. If the average open case reserve per claim at twelve months of development has been rising over recent accident years while claim frequency is flat, the claims team is posting higher initial reserves. That is not inherently bad, but the actuary needs to know it happened.
-
Closed-without-payment rate. If the percentage of claims that close without any payment is trending down, it often means files that used to close quickly are now staying open longer or settling for small amounts. That shifts the development pattern.
The actuary’s remedy is the Berquist-Sherman case-reserve adjustment (Friedland, p. 283). The technique adjusts historical incurred claims to reflect current reserving levels, stripping out the effect of the practice change so the development pattern is on a consistent basis. The 1977 Berquist and Sherman paper remains the most cited diagnostic paper in casualty reserving, and the adjustment it describes is the standard tool for this problem.
Your operational remedy is more direct: ask your claims team three questions. Has the case reserving philosophy changed in the last two years? Are adjusters seeing claim types they have not seen before? Have caseloads grown to the point where file reviews are getting compressed? The answers tell you whether a Berquist-Sherman adjustment is the right tool, or whether the underlying issue is a staffing problem that no actuarial technique can fix.
Concrete action: Ask your actuary for a closed-claim-only triangle alongside the standard incurred triangle. If the closed-claim triangle shows substantially less adverse development than the all-claims triangle, the problem is almost certainly in open case reserves, not in ultimate claims cost. That single comparison will tell you more about what is driving your number than twenty pages of selected development factors.
Diagnostic 2: Are your payment patterns shifting?
Friedland’s failure-mode table explicitly flags both settlement speed-up and settlement slowdown as conditions that break the paid chain ladder (Friedland, Section 4). The paid chain ladder is calibrated to a historical payment velocity. When that velocity changes, the method projects the wrong ultimate.
Payment patterns shift for operational reasons, not actuarial ones. A TPA transition almost always disrupts payment velocity for six to twelve months. A claims management initiative pushing faster settlements will compress the paid pattern. A change in a state’s workers compensation medical fee schedule can shift the timing and size of medical payments. A legislative change to indemnity benefit duration, the kind NCCI tracks in its research on temporary disability duration, can extend or shorten the payment tail for an entire line.
How to spot it. Two signals:
-
Paid-to-paid age-to-age factors trending in one direction over recent diagonals. Pull the most recent three to four diagonals of your paid triangle and compare the 12-to-24 and 24-to-36 development factors. If they are consistently rising, payments are coming in faster or larger than historical norms. If they are consistently falling, payments are decelerating. A single outlier diagonal is noise. Three in a row is a trend.
-
Average payment per closed claim drifting. If average closure cost is rising but claim counts are flat, individual claims are settling for more. If average cost is flat but the number of closures per quarter is rising, claims are closing faster. Both reshape the paid development pattern.
The actuary’s remedy for a settlement-rate shift is the Berquist-Sherman disposal-rate adjustment, which restates the paid triangle on a constant-disposal-rate basis (Friedland, p. 287). For a severity shift, the actuary may need to separate frequency and severity trends and project them independently.
Concrete action: Ask your claims team what has changed in the last eighteen months in how claims close. TPA transitions, settlement authority changes, litigation management changes, fee schedule updates. Write the answer down and hand it to your actuary before the next review. An actuary who knows a TPA transition happened in Q3 can handle it. An actuary who discovers it by staring at a triangle diagonal is working backwards from a symptom, which is slower and less reliable.
Diagnostic 3: Has your claim mix shifted?
This is the hardest driver to see from a triangle and the most important to catch early. Friedland flags change in mix toward higher-severity claims as a condition that breaks the paid chain ladder because small claims close first and large claims stay open, which means the development pattern in a blended triangle is driven by whatever the current proportion of large to small claims happens to be (Friedland, Section 4).
Mix shifts happen when your business changes. A manufacturer that adds a product line with chemical exposure risk does not just add claims; it adds a different kind of claim with a longer tail and higher average severity. A trucking company that expands from regional to long-haul routes changes its accident profile. A public entity that increases police or fire headcount increases its exposure to the claim types with the longest development patterns. A hospital system that opens a higher-acuity surgical unit shifts its malpractice exposure mix.
None of these changes show up naturally in an aggregate loss triangle. The triangle treats every accident-year dollar the same regardless of where it came from. It cannot tell you that sixty percent of your development is concentrated in one location or claim type, because it was never asked to.
How to spot it. This diagnostic requires data the triangle does not contain. You need claim counts and average severity by line of business, location, or claim type over time. If one segment’s severity is growing substantially faster than the others, or if one segment’s share of total claims is rising, your mix has shifted and your aggregate triangle is blending old-mix development patterns with new-mix exposure.
Friedland references Berquist and Sherman’s general principle that the actuary should reorganize data when conditions change: policy year for accident year on limit changes, accident quarter for accident year on rapid growth (Friedland, p. 283). For a mix shift, that means segmenting: building separate triangles for distinct lines, locations, or claim types, and projecting each independently. A blended triangle on a changing book will produce the wrong answer. Not because the math is wrong, but because the input is a mixture of things that develop differently.
Concrete action: Ask your actuary to segment the most recent triangle by line of business, location, or claim type and show you where development is concentrated. If eighty percent of your adverse development is coming from twenty percent of your exposure base, you know exactly where to focus. That request changes the conversation from “reserves went up” to “reserves went up because of this specific part of the book,” which is the only version that leads to operational action.
The leading indicators your TPA already has but your actuary may not see
The diagnostics above are retrospective. They tell you what happened after the development materialized. The metrics below are prospective. They predict reserve development before it shows up in a triangle, and your TPA already tracks most of them.
Average days to first report. The gap between the date of loss and the date the claim is first reported to your TPA. When this number rises, it means either injured workers are waiting longer to report, or the internal reporting chain has slowed down. Either way, late-reported claims are more expensive on average than promptly reported claims, and an increase in reporting lag predicts adverse pure IBNR development in the next valuation.
Average days from first report to first reserve. Once the claim reaches the TPA, how long before an adjuster sets an initial case reserve? A rising number here means the claims team is slower to assess, which delays the information the incurred chain ladder depends on and correlates with larger ultimate claim costs.
Percentage of claims with attorney representation. Attorney-represented claims cost more and take longer to close. If attorney involvement is trending up across your book, your severity and duration patterns are both shifting, and historical development factors will understate ultimate losses. This is one of the strongest single predictors of adverse development in general liability and workers compensation programs.
Average medical-only claim duration. In workers compensation, medical-only claims are the high-frequency, low-severity base of the triangle. When their average duration starts creeping up, it often signals that return-to-work coordination has slipped. Individually these claims are small, but collectively their development pattern anchors the early columns of the triangle, and a duration shift changes the shape of the overall pattern.
Frequency of large-loss notices. How often is the TPA flagging claims that may breach your retention or stop-loss attachment? A rising frequency of large-loss notices predicts adverse development on the specific claims that carry the most tail risk. Your actuary may be capping large claims in the triangle, but if the cap is based on last year’s threshold and this year’s losses are routinely reaching it, the cap is stale.
None of these five metrics appear in a standard reserve study. The actuary works from a triangle built from transaction dates and dollar amounts. The TPA tracks these operational metrics, but unless someone delivers them to the actuary, they sit in dashboards that nobody connects to the reserve number.
Concrete action: Ask your TPA for a quarterly report on these five metrics, trended over the last eight quarters. You do not need the actuary to request it. You are looking for trends that change direction. A metric stable for two years that moves in the last three quarters is a signal. Hand the trend report to your actuary before the next review and ask whether it changes anything about the development assumptions.
What to do with what you find: the conversation with your actuary
Diagnostics are only useful if they change the questions you ask. Once you have identified which of the three drivers is most likely in play, you walk into the actuarial review meeting with a specific question instead of a general concern.
If case reserve adequacy is the issue: “Our case reserves have been strengthening over the last two years. Can you show me the incurred chain ladder result side by side with a Berquist-Sherman case-adjusted version, so we can see how much of the development is practice change versus true loss emergence?”
If payment patterns are shifting: “We transitioned TPAs eighteen months ago and settlement speed has changed. Can you run the paid chain ladder on both the raw triangle and a constant-disposal-rate basis per Berquist-Sherman, and show us the difference?”
If claim mix is the driver: “We expanded into long-haul trucking two years ago. Can you build a separate triangle for long-haul versus regional and project them independently, so we can see whether the aggregate development is being driven by the new exposure?”
Each of these questions names a specific technique. Berquist-Sherman is the diagnostic adjustment framework that Friedland references extensively (Friedland, pp. 283, 287), and it is the standard the actuary is trained on. Knowing the name and asking for it by name changes the dynamic of the conversation. You are no longer a consumer of a number; you are a participant in the analysis.
One important boundary: you are not asking the actuary to lower the number. You are asking whether the number reflects a real change in loss costs or an artifact of changing conditions. If the diagnostic confirms that your losses are genuinely getting more expensive, that is valuable too, because it tells you where to focus risk management effort rather than spreading attention across the whole book.
What this does and does not do for IBNR
An honest section. These diagnostics tell you why your historical development looks the way it does. They help you and your actuary produce a more accurate estimate of what you owe today. They do not directly lower future IBNR.
Lowering IBNR runs through risk management actions: return-to-work programs that reduce indemnity duration, large-loss review protocols that catch runaway claims early, claims-handling discipline that keeps initial reserves accurate and files moving, and exposure-base changes that reduce the frequency or severity of losses before they enter the triangle. Future articles in this series will cover specific risk management interventions and their measurable impact on reserve development.
What diagnostics do accomplish is equally important. They remove the fog from the reserve conversation. A CFO who knows that adverse development is concentrated in one business unit, driven by a TPA transition, or caused by a shift in case reserving practice is in a fundamentally different position from one who just knows the number went up. The first CFO can act. The second can only worry.
Diagnostics also protect you from a costly mistake: overreacting to adverse development that is an artifact rather than a real change in loss costs. If your case reserves strengthened because your claims team got more conservative, and your actuary did not adjust for it, the incurred chain ladder will overstate your ultimate losses. You will book a reserve that is too high and potentially make pricing or operational decisions based on a loss trend that is not real. The diagnostic catches the artifact before it becomes a bad decision.
Three concrete actions for Monday morning
1. Ask your TPA for the leading indicators. Request a quarterly trend report covering average days to first report, days to first reserve, attorney representation rate, medical-only claim duration, and large-loss notice frequency. Ask for the last eight quarters so you can see direction of travel. This data already exists in the TPA’s system. Getting it is a phone call, not a project.
2. Identify which driver is most likely in play. Before your next actuarial review, ask yourself: is the development concentrated in recent accident years (case reserve issue), spread across diagonals (payment pattern issue), or concentrated in one part of the book (mix issue)? You do not need to be precise. You need a hypothesis.
3. Walk into the review with the specific question. If case adequacy is the issue, ask for a Berquist-Sherman case-adjusted comparison. If payment patterns shifted, ask for a constant-disposal-rate analysis. If mix changed, ask for a segmented triangle. Cite the technique by name. The actuary will either confirm your hypothesis or explain why a different driver is more likely, and either outcome is more valuable than sitting through a forty-page triangle presentation without a question.
If you want a second set of eyes on any of this, the independent review note explains why an independent actuarial assessment can add clarity, especially when the relationship between the reserve study, the TPA, and the program’s own operations makes it hard to diagnose from inside.
For readers who arrived here without the foundational concept, the plain-English IBNR explainer covers the building blocks this article builds on.