Self-Insured & Captive

Commercial Auto and Fleet IBNR for Self-Insured Trucking, Delivery, and Transit

How actuaries estimate unpaid auto liability claims for self-insured fleets, why bodily injury severity dominates the reserve, and what to require in the report.

By Sam · April 2026 · 13 min read

If your organization operates a commercial fleet (whether long-haul trucking rigs, last-mile delivery vans, transit buses, or service vehicles) and self-insures the auto liability, your balance sheet carries a liability for claims that have already occurred but have not yet been fully paid. The most volatile piece of that liability is IBNR: incurred but not reported, which in actuarial practice includes both claims nobody has reported yet (pure IBNR) and future development on claims already in your adjuster’s inventory (IBNER). Together, those two components are what actuaries call broad IBNR, and for commercial auto liability, they are dominated by bodily injury claims that can take years to resolve and, increasingly, by nuclear verdicts that distort historical development patterns.

This article explains how actuaries estimate that number for self-insured commercial fleets, what makes the line different from other casualty lines, where the standard methods fail, and what you should require in the report. It is written for the fleet risk manager, CFO, or captive manager who receives a reserve study and needs to evaluate it.

Why commercial auto is a two-tail problem

Commercial auto liability is not one line. It is at least two: bodily injury (BI) and physical damage (PD), with fundamentally different reserving characteristics.

Bodily injury has a medium-to-long tail. Serious BI claims involving catastrophic injuries, wrongful death, or multi-party accidents can take three to seven years to close, longer if litigation reaches appeal. Case outstanding on these files changes as medical treatment continues, settlement negotiations evolve, and litigation risk is reassessed. The development pattern is driven by severity: how large each claim ultimately settles, not how many claims are filed.

Physical damage has a short tail. Most PD claims close within twelve to eighteen months. Payments are predictable, case reserves are set quickly from repair estimates, and the development pattern is driven by claim count and closure speed. PD is, in reserving terms, a relatively simple line.

The consequence for the actuary (and for you) is that the two coverages require different methods, different development patterns, and different diagnostic attention. A report that combines BI and PD into a single triangle is hiding the coverage that matters most behind the coverage that matters least. If your report does not separate BI from PD, that is the first question to ask.

The exposure base

Friedland identifies vehicle count or miles driven as the standard exposure base for commercial automobile liability (Friedland, p. 132). That means the expected-claims anchor in a Bornhuetter-Ferguson or expected-claims analysis is built from fleet size or utilization times an expected claim rate per unit of exposure.

For a self-insured fleet, the exposure base matters more than it might for a carrier because fleet size can change quickly. An acquisition that doubles your tractor count, a route expansion that increases miles driven by 40%, or a reduction in force that shrinks the fleet all change the denominator of the expected-claims calculation. If the actuary is using last year’s expected claim ratio (what the industry calls a loss ratio; Friedland uses “claim ratio” because she reserves “loss” for non-claims economic concepts) applied to this year’s exposure without adjusting for the mix of vehicle classes (long-haul versus local, heavy truck versus light commercial), the expected-claims anchor is wrong, and any method that relies on it (Bornhuetter-Ferguson, expected claims, Cape Cod) inherits the error.

The arithmetic in plain language

The core equation is the same one that applies to every casualty line:

Ultimate claims = paid to date + case outstanding + IBNR

The IBNR explainer walks through this in detail. For commercial auto, the actuary typically builds separate triangles for BI and PD. Consider a self-insured trucking fleet with the following BI experience:

Your 2024 accident year shows $2.8 million in reported BI losses (paid plus case outstanding) as of December 2025, at 24 months of development. The actuary’s selected development pattern says that BI losses at 24 months are historically about 55% of ultimate. The cumulative development factor (CDF, also called an LDF) is 1.818.

Projected ultimate = $2.8 million x 1.818 = $5.09 million

The broad IBNR for this year is $5.09 million minus $2.8 million = $2.29 million, or about 45% of the projected ultimate. At 12 months, the CDF might be 3.00 or higher, meaning more than two-thirds of the projected ultimate is an actuarial estimate rather than an observed number.

For PD, the same fleet might show $1.2 million in reported losses at 24 months with a CDF of 1.02, producing a projected ultimate of $1.22 million and an IBNR of just $24,000. The PD reserve is almost entirely observed; the BI reserve is almost entirely estimated. That asymmetry is why the BI triangle deserves most of the analytic attention.

Nuclear verdicts and the severity problem

The defining trend in commercial auto reserving over the past decade is the acceleration of BI severity, driven in significant part by nuclear verdicts: jury awards that exceed historical norms by multiples, sometimes exceeding $10 million or $50 million on a single claim. The effect on a self-insured fleet’s reserve is twofold.

First, severity trend dominates frequency trend. Even if your fleet’s accident frequency is stable or declining (better driver training, telematics, collision avoidance systems), the severity of the claims that do occur is rising faster than frequency is falling. A report that applies a combined frequency-severity trend of 3% when frequency is declining at 2% and severity is rising at 8% is understating the true trend by a material margin.

Second, nuclear verdicts distort historical development patterns. A single $15 million verdict in an accident year that previously had $3 million in reported losses does not look like normal development; it looks like a break in the pattern. The age-to-age factor for that maturity interval spikes, and if the actuary’s selected factor includes that spike, the leveraged factor is applied to every other accident year at the same maturity, potentially overstating them. If the actuary excludes the spike, the leveraged factor may understate the years where a similar verdict is plausible. There is no clean answer, only judgment and documentation.

Method selection: frequency-severity for BI, chain ladder for PD

The five core methods overview describes method selection in general terms. For commercial auto, the typical approach differs between coverages.

Bodily injury. Frequency-severity analysis is often preferred for BI, particularly for the most recent accident years, because it separates the claim count projection from the average severity projection. This is valuable when severity is volatile (as it is in the nuclear-verdict environment) because it isolates the severity assumption where the actuary can apply judgment, select a trend, and test sensitivity without letting a single large claim contaminate the entire development pattern. Bornhuetter-Ferguson is the standard alternative for recent years, and the chain ladder is typically reserved for older years where the development pattern has stabilized.

The leveraged-factor risk for BI is real. A self-insured fleet with 30 BI claims in the most recent accident year and a 12-month CDF of 3.50 faces a situation where a single $200,000 settlement added to the most recent year produces a $700,000 swing in the projected ultimate. Frequency- severity avoids this by anchoring the severity estimate to a trend line rather than riding the volatile development factor forward.

Physical damage. Chain ladder on reported losses is usually sufficient for PD. The tail is short, the claims are homogeneous, and the development pattern is stable. The actuary may not even run alternative methods for PD, and that is generally acceptable.

Salvage and subrogation

PD claims routinely produce salvage (recovery from the sale of damaged vehicles) and subrogation (recovery from at-fault third parties). These recoveries are material for PD and can produce development factors below 1.000 at later maturities (Friedland, p. 329), meaning reported losses actually decline as recoveries come in. A report that does not account for salvage and subrogation in the PD development pattern will overstate the PD ultimate. If the PD triangle shows no factors below 1.000 at any maturity, ask whether S&S recoveries are included or netted separately.

What breaks the estimate

Every reserving method makes assumptions, and when those assumptions fail, the method still produces a number; it is just the wrong number. Commercial auto programs are especially prone to three categories of change that invalidate the historical development pattern.

Fleet composition and utilization changes

Adding a new operating division, acquiring a competitor’s fleet, or shifting from owner-operators to company drivers changes both the exposure base and the expected severity profile. Long-haul operations produce different BI severity distributions than local delivery routes. If the actuary applies a single expected claim rate across the entire fleet without adjusting for composition, the expected-claims anchor is distorted for every accident year that straddles the change.

TPA transitions and claims handling shifts

When a self-insured fleet changes its third-party administrator, the new TPA typically reviews open files and resets case reserves (also called case outstanding in actuarial terminology) to its own adequacy standards. The case reserve strengthening article explains how this distorts the reported triangle. For BI claims with long settlement timelines, a TPA change mid-stream can affect multiple open accident years simultaneously, producing a calendar-year adequacy shift that the unadjusted chain ladder misreads as adverse development.

Litigation environment changes

Changes in venue rules, tort reform, or jury attitudes toward commercial carriers shift the BI severity distribution in ways that do not appear in historical triangles until verdicts arrive. A state that removes caps on non-economic damages, or a jurisdiction that becomes plaintiff- friendly for trucking cases, changes the prospective severity trend without changing the historical one. The actuary needs to know about these shifts and explain how they are reflected in the trend selection.

The self-insured wrinkle: thin data, big claims

Self-insured fleets face the same thin-data problem described in the workers compensation article, but the severity concentration is often worse. A carrier writing commercial auto might have thousands of BI claims per accident year. A self-insured fleet with 500 vehicles might generate 20 to 40 BI claims per year, of which two or three are serious enough to drive most of the loss dollars. When you build a loss development triangle from that volume, a single large settlement in one maturity interval can dominate the age-to-age factor for that column.

The standard mitigations are the same ones that apply to any thin-data self-insured program: Bornhuetter-Ferguson for recent years (which caps the influence of actual data at a credibility weight that increases with maturity), frequency-severity analysis (which isolates the large-loss volatility in the severity component), and limited-loss analysis (which caps individual claims at the per-occurrence SIR or retention and projects the capped triangle separately from the excess layer). The limited triangle has less volatility because large claims are truncated, and the excess layer can be reserved using expected-claims methods or industry benchmarks.

If your program has a per-occurrence retention and the report does not show a limited-loss analysis, ask why. The self-insurance structures article explains how the retention structure changes the reserving problem.

What a buyer should ask their actuary

Are BI and PD separated in the analysis? If not, the BI reserve is hidden behind PD, and you cannot evaluate the assumption that matters most.
What method was used for BI for each accident year, and why? Recent years should show Bornhuetter-Ferguson, frequency-severity, or both. If the chain ladder drives the most recent year, ask what justifies trusting the leveraged CDF with your claim volume.
What severity trend is selected for BI, and what is the basis? The severity trend for commercial auto BI should be documented from program experience or an appropriate industry benchmark, separated from frequency trend. A blended “loss trend” without decomposition hides the severity assumption that drives the recent-year estimate.
How are nuclear verdicts or large losses handled in the development pattern? The actuary should disclose whether large losses are capped, excluded, or included in the triangle, and how excess losses are treated in the projection.
Are salvage and subrogation reflected in the PD development pattern? If the PD factors at later maturities are not below 1.000, recoveries may be missing from the analysis.
Is the estimate a central estimate, a management best estimate, or something else? ASOP 43 defines the actuarial central estimate as an expected value over reasonably possible outcomes. Know which one you are getting.

What to require in documentation

The documentation checklist consolidates professional-standards requirements into a single reference. For commercial auto specifically, the report should include:

Separate BI and PD triangles at annual or semiannual evaluation intervals, with enough history to observe the full BI development pattern through settlement.
Method selection by accident year and coverage with stated rationale, including a comparison of chain ladder, Bornhuetter- Ferguson, and frequency-severity results for BI.
Frequency and severity trend selections documented separately, with source (program experience, industry benchmark, or blend) and the combined annual trend applied to the expected-claims anchor.
Large-loss treatment disclosing the threshold, the number and amount of large losses by accident year, and how they are handled in each method (included, excluded, capped, or developed separately).
Salvage and subrogation treatment for PD, including whether recoveries are netted in the triangle or projected separately, and the basis for the recovery assumption.
Exposure base reconciliation confirming that vehicle counts or miles driven by accident year tie to fleet records, with adjustments for fleet composition changes.
Operational change disclosure documenting any changes in fleet size, route structure, driver training programs, telematics implementation, claims handling, or TPA transitions during the experience period.
Sensitivity analysis showing how the BI reserve changes under alternative severity trend, CDF, or expected-claims assumptions for the two or three most recent accident years.

If the report does not include these elements, send the list back and ask for the missing sections. A report that projects BI and PD together, selects a single blended trend, and provides no sensitivity analysis is not giving you the information you need to evaluate the number on your balance sheet.