15,000 PFAS Lawsuits Expose GL Reserving Gaps for Actuaries

As of April 2026, 15,222 personal injury PFAS lawsuits sit in MDL 2873 before Judge Richard Gergel in the U.S. District Court for the District of South Carolina. Out of 19,788 total cases filed to date, the remaining docket comprises firefighters, military veterans, and community residents alleging kidney cancer, testicular cancer, thyroid disease, and ulcerative colitis from per- and polyfluoroalkyl substance exposure. Bellwether trials, postponed from October 2025, are expected in 2026, with a pool of six cases narrowed to three kidney cancer and three testicular cancer claims. Settlement fund deadlines for the existing $10.3 billion 3M water-system agreement fall between June and August 2026, adding urgency to carrier exposure quantification.

The financial scale is not speculative. 3M committed $10.3 billion over 13 years to remediate public water systems. DuPont, Chemours, and Corteva settled for $1.185 billion with 300 local water districts. New Jersey secured an $875 million resolution in August 2025 for contamination at four former manufacturing sites. Verisk's Arium liability modeling group projects total ground-up PFAS litigation losses at $120 to $165 billion, placing the exposure in the same order of magnitude as asbestos. Carrier Management has framed the risk as "the next asbestos," yet Milliman research confirms that dedicated PFAS reserves remain uncommon across the industry.

For casualty actuaries, the gap between the magnitude of projected losses and the absence of structured reserve programs creates a professional risk. This article examines why standard reserving techniques fail for PFAS, outlines the exposure-based framework that Milliman and Praedicat recommend as a replacement, and walks through the coverage allocation mechanics that make PFAS IBNR estimation fundamentally different from conventional GL development.

Why Traditional Reserving Methods Fail for PFAS

Chain-ladder development, Bornhuetter-Ferguson, and Cape Cod methods all depend on one core assumption: that historical claim emergence patterns will repeat in future periods. For PFAS, that assumption fails at every level.

Non-stationary claim emergence. The MDL currently tracks four disease categories: kidney cancer, testicular cancer, thyroid disease, and ulcerative colitis. Each has a different latency period, a different epidemiological evidence base, and a different litigation maturity curve. More than 10,000 peer-reviewed studies have examined PFAS-health associations over the past two decades, and research continues to identify new conditions. When the plaintiff population itself is expanding as science matures, historical development factors calculated from earlier cohorts systematically understate ultimate losses. A loss development triangle constructed from 2015 data cannot anticipate the kidney cancer claims that emerged in force after 2020, and it cannot predict what new disease link the next epidemiological study will establish.

No stable tail factor. In conventional GL reserving, actuaries select a tail factor to project beyond the last observable development period, typically anchoring to industry benchmarks or prior accident-year experience. PFAS claims lack both. The exposure has no historical analog with comparable latency, toxicological diversity, and regulatory uncertainty. Asbestos comparisons are instructive but imperfect: asbestos claim emergence plateaued after mesothelioma science stabilized, while PFAS epidemiology remains in active development. Applying an asbestos-derived tail to PFAS would embed a maturity assumption that the science does not support.

Regulatory fragmentation distorts severity. The EPA finalized a 4 parts per trillion maximum contaminant level (MCL) for PFOA and PFOS in April 2024, with water systems required to comply by 2029 (extended to 2031 under a May 2026 proposal). But state-level MCL thresholds range from 4 ppt to 70 ppt, creating a patchwork of remediation cost obligations. An insured site in a state with a 4 ppt standard faces remediation costs that can be multiples of the same site in a 70 ppt jurisdiction. Since the chain-ladder method assumes severity patterns are consistent across the development period, this regulatory heterogeneity introduces systematic bias into any aggregate development approach. Actuaries would need to model cleanup liability under multiple regulatory scenarios rather than selecting a single severity trend.

The Exposure-Based Alternative: Milliman and Praedicat Framework

Milliman and Praedicat have published research demonstrating that exposure-based stochastic modeling is the appropriate IBNR framework for PFAS. The methodology starts from the exposed population rather than historical claim counts and builds upward through a four-stage process.

Stage 1: Map the insured exposure universe. Using EPA contamination data, the actuary identifies geographic areas with documented PFAS contamination and cross-references them against the insured portfolio's policy locations. For a GL book, this means geocoding each insured premises against EPA National Priorities List sites, military base AFFF usage records, and state-level contamination databases. The output is an estimate of the number of individuals exposed to PFAS through each insured's operations or premises over each policy year.

Stage 2: Apply disease incidence rates. From the body of peer-reviewed epidemiological research, the actuary selects dose-response relationships for each disease category. The C8 Science Panel studies, which examined 69,000 residents near DuPont's Washington Works facility, established relative risk ratios for kidney cancer (1.58x), testicular cancer (1.54x), thyroid disease (1.25x), and ulcerative colitis (1.76x) in high-exposure populations. These ratios, applied to baseline population incidence rates, produce expected disease counts per exposed cohort.

Stage 3: Estimate claim filing propensity. Not every injured individual files a lawsuit. The actuary calibrates a litigation-maturity curve using data from analogous mass torts. Asbestos claim filing propensity peaked roughly 20 to 30 years after peak exposure for mesothelioma claimants. Talc litigation followed a compressed timeline after media coverage accelerated awareness. For PFAS, the filing propensity function must account for the dual accelerants of MDL consolidation (which reduces filing friction) and ongoing epidemiological publications (which increase both awareness and legal causation strength). Praedicat's Nekomodel stochastically simulates the future path of scientific literature to project when filing propensity inflection points occur.

Stage 4: Run stochastic severity simulations. Bellwether trial outcomes will establish initial severity benchmarks. Current projections estimate $200,000 to $600,000 per Tier 1 claimant (kidney or testicular cancer with occupational AFFF exposure), with lower tiers for thyroid disease and ulcerative colitis. The stochastic model draws from a severity distribution calibrated to these benchmarks, adjusted for jurisdiction, disease category, and exposure duration. Praedicat's model runs thousands of simulations to produce a distribution of ultimate losses at the account and portfolio levels. Milliman's proprietary model follows a similar stochastic structure, calibrated with the latest litigation data and settlement benchmarks.

The output is not a point estimate but a probability distribution of ultimate PFAS losses, with percentile selections available for carried reserves, actuarial opinions, and reinsurance cession calculations.

Coverage Allocation: Why Policy-Year Analysis Is Unavoidable

PFAS exposure creates a coverage allocation problem that is absent from most standard GL reserving. The complication is temporal: PFAS contamination occurred over decades, policies changed form over that period, and the trigger theory that a court applies determines which policy years respond.

The pollution exclusion timeline. The insurance industry introduced the "sudden and accidental" pollution exclusion around 1973 and replaced it with the absolute pollution exclusion in approximately 1986. Pre-1973 occurrence-form GL policies contain no pollution exclusion and respond to PFAS claims without coverage dispute on those grounds. Policies from 1973 to 1986 may respond depending on whether the contamination qualifies as "sudden and accidental" under state law (courts are split). Post-1986 policies with absolute pollution exclusions generally do not respond to PFAS environmental claims, though coverage arguments persist in some jurisdictions where courts have narrowly construed the exclusion. Current-year GL policies increasingly include explicit PFAS exclusions endorsed by ISO and surplus lines carriers.

The practical consequence: PFAS IBNR concentrates in older policy years where no exclusion existed, while current underwriting years carry minimal PFAS exposure. An actuary cannot estimate aggregate PFAS IBNR without performing a retrospective policy-form analysis across decades of underwriting history to identify which policy years have responsive coverage.

Worked Example: Allocating IBNR Across Policy Years

Consider an insurer with a GL book that includes industrial manufacturing accounts written continuously from 1965 through 2026. The stochastic model produces a mean ultimate PFAS loss of $50 million for this portfolio segment. The actuary must now allocate that $50 million across policy years. Three allocation methods are in common use for long-tail environmental claims:

Method	Allocation Logic	Effect on Pre-1973 Years
Pro rata (time on risk)	Spreads loss equally across all triggered policy years proportional to time	Moderate: shares loss across all years, diluting concentration in any single year
All sums (joint and several)	Policyholder selects the policy year with the broadest coverage and deepest limits	Heavy: concentrates loss in pre-exclusion years with full coverage response
Injury-in-fact	Allocates to the specific year(s) when bodily injury actually occurred	Variable: depends on when disease manifested, which may be decades after exposure

Source: Hinshaw & Culbertson PFAS Insurance Coverage Primer; Pillsbury PFAS liability coverage analysis.

Under the all-sums approach, which several jurisdictions apply, the $50 million would concentrate in the pre-1973 policy years that lack any pollution exclusion. Under pro rata allocation across a 1965 to 2000 trigger period, each year would absorb roughly $1.4 million. Under injury-in-fact, the allocation depends on when each claimant's disease manifested, which could span from the 1980s through the 2030s as latent conditions emerge.

The choice of allocation method is not actuarial; it is determined by the governing jurisdiction's case law. But the actuary must model IBNR under each plausible method and present the results as a range. A single-point IBNR estimate that assumes one allocation method across a multi-state book will be wrong for every state that applies a different trigger theory.

Why $120 to $165 Billion Requires Carrier-Level Analysis

Verisk's Arium model projects total ground-up PFAS losses at $120 to $165 billion before insurance coverage terms and conditions. That headline figure is useful for framing industry-level magnitude, but it is not directly usable for individual carrier reserving for three reasons.

First, the ground-up figure precedes coverage application. After applying pollution exclusions, policy limits, retentions, and coverage allocation, the insured industry loss will be a fraction of the ground-up total. The fraction varies enormously by carrier depending on vintage of book, policy forms used, and jurisdictional mix.

Second, PFAS exposure is concentrated in specific industries. Arium's model identifies PFAS liability rippling through interconnected supply chains, from chemical manufacturers to firefighting foam distributors to military installations to municipal water systems. A carrier with heavy manufacturing GL exposure faces a fundamentally different PFAS liability than one whose book is concentrated in professional services.

Third, the loss projection is sensitivity-dependent. The range between $120 billion and $165 billion reflects assumptions about claim filing rates, disease-category expansion, regulatory cleanup cost escalation, and litigation duration. Each carrier must stress-test these assumptions against its own exposure profile rather than applying an industry average.

This is why Milliman and Praedicat emphasize account-level and portfolio-level modeling over market-level benchmarks. An actuary who sets a PFAS reserve based on a market share allocation of the $120 to $165 billion figure will produce an estimate that is neither defensible nor useful for management decision-making.

Practical Steps for Casualty Actuaries

From tracking how carriers have responded to PFAS exposure over the past two years, several patterns have emerged in best-practice reserving approaches:

1. Perform a policy-form inventory. Before any modeling begins, the actuary needs a complete accounting of which policy years used occurrence forms without pollution exclusions, which included sudden-and-accidental language, and which carried absolute exclusions. For many carriers, this means retrieving archived policy forms from the 1960s through the 1980s, a process that can take months for books acquired through M&A.

2. Build exposure maps, not triangles. Cross-reference the insured portfolio against EPA contamination databases, military AFFF site lists, and state environmental agency records. The goal is to identify which accounts had insured operations in proximity to documented PFAS contamination during responsive policy periods.

3. Adopt scenario-based severity assumptions. Until bellwether trials produce actual verdicts, severity remains assumption-driven. Model at least three scenarios: a low scenario using current water-system settlement per-claimant averages (roughly $40,000 to $80,000), a base scenario using projected Tier 1 personal injury values ($200,000 to $600,000 for cancer claims), and a high scenario incorporating nuclear verdict potential in plaintiff-friendly jurisdictions.

4. Model regulatory cost escalation explicitly. With MCL thresholds ranging from 4 ppt to 70 ppt across states and the EPA potentially rescinding MCLs for four PFAS compounds while maintaining the PFOA/PFOS standard, remediation cost projections must include scenario branches for regulatory tightening, maintenance of current standards, and partial rollback.

5. Disclose methodology and uncertainty. ASOP No. 43 (Unpaid Claim Estimates) requires the actuary to disclose the methods used and the rationale for selecting them. For PFAS, the rationale for departing from standard development methods is well-supported by the non-stationary claim emergence pattern. The actuarial opinion should explicitly state that traditional methods were considered and rejected, and that the exposure-based approach was selected because PFAS violates the stationarity assumption underlying loss development techniques.

Why This Matters for Pricing Actuaries

The reserving problem feeds directly into pricing. Current-year GL policies increasingly exclude PFAS through ISO and surplus lines endorsements, but the long-tail nature of the exposure means that pricing actuaries must also consider how PFAS IBNR development on older years affects the overall loss ratio and reserve position that flows into rate indications.

If a carrier's aggregate GL reserve position deteriorates because of PFAS IBNR recognition, that adverse prior-year development shows up in the Schedule P data that pricing actuaries use for rate filings. Carriers that have not yet established PFAS reserves may show artificially favorable loss ratios in historical experience periods, leading to rate indications that understate the true cost of the GL book. This is a variant of the same social inflation distortion in casualty loss development that affects GL and excess casualty pricing, compounded by the latent, multi-decade character of PFAS claims.

The $120 to $165 billion industry projection, combined with the concentration of IBNR in pre-exclusion policy years, means that PFAS reserve recognition will likely produce a multi-year drag on casualty reserve adequacy metrics across the industry. Actuaries tracking the commercial auto reserve gap and CNA's Q1 casualty reserve charge should view PFAS as an additional source of adverse development risk that sits outside the standard loss development framework entirely.