How Social Inflation Is Distorting Casualty Loss Development Factors

From reviewing casualty loss triangles across multiple GL and excess casualty books over the past two years, one pattern keeps emerging: the age-to-age factors at the 36-to-48 and 48-to-60 month maturities on recent diagonals are running persistently above their five-year weighted averages. The gap is not random noise. It reflects a structural shift in how casualty claims develop, driven by nuclear verdict acceleration, third-party litigation funding, and changing juror attitudes that collectively fall under the heading of social inflation.

The industry data confirms what individual triangles are showing. An Insurance Thought Leadership analysis published in January 2026 found that 47% of companies examined showed adverse reserve development for accident years 2022 and prior. CNA Financial’s Q1 2026 results drove the point home: a loss ratio of 71.8% (up from 67.8%), with 4.1 points of unfavorable prior-period development concentrated in excess casualty and professional E&O. AM Best’s February 2026 assessment identified a $9 billion net reserve deficiency at year-end 2024, with Products/Other Liability posting a combined ratio of 108.0. The CAS and Insurance Information Institute joint report estimated $231.6 billion to $281.2 billion in cumulative excess liability losses over the prior decade, far exceeding what economic inflation alone can explain.

For pricing actuaries, the question is no longer whether social inflation is real. It is how to adjust your loss development factors, tail selections, and excess loss loads so that rate indications reflect the losses that are actually coming.

47%

Carriers With Adverse Reserve Development

$9B

Industry Net Reserve Deficiency (YE 2024)

108.0

Products/Other Liability Combined Ratio

$281B

Cumulative Excess Liability Losses (10-Year)

Detecting Diagonal Distortion in Your Triangle

The standard chain-ladder method assumes that future loss development will mirror historical patterns. Social inflation violates this assumption by introducing a calendar-year effect: claims from accident years 2018 through 2022 are developing at persistently higher link ratios than historical norms, particularly at the 36-to-48 and 48-to-60 month maturities where litigation outcomes begin to materialize.

The diagnostic is straightforward. Pull your incurred loss development triangle and compute age-to-age factors by accident year. Then compare the most recent three calendar-year diagonals against the all-year volume-weighted average at each maturity. A persistent uptick of 2% or more at the medium-tail maturities (36-60 months) that does not revert to the mean on the next diagonal is a social inflation signal, not statistical noise.

Consider a simplified GL occurrence triangle. The table below illustrates the pattern:

Maturity	All-Year Wtd Avg	Latest 3 Diagonals	Difference
12-24	2.050	2.070	+0.020
24-36	1.420	1.435	+0.015
36-48	1.180	1.205	+0.025
48-60	1.090	1.110	+0.020
60-72	1.045	1.050	+0.005
72-84	1.025	1.028	+0.003

The highlighted 36-48 and 48-60 maturities show the steepest divergence. This is where litigation outcomes, settlement negotiations, and nuclear verdicts hit the incurred loss development pattern. If you select LDFs using the all-year weighted average without adjusting for this diagonal shift, you will systematically understate ultimate losses for the accident years still developing through these maturities.

Berquist-Sherman: Stripping Out Reserve Adequacy Changes

Before attributing all of the diagonal uptick to social inflation, pricing actuaries need to isolate the effect from a confounding variable: changes in case reserve adequacy. If your claims department strengthened case reserves during the same calendar period (as many did in response to adverse development), the incurred loss triangle will show higher age-to-age factors even if the underlying claim costs did not change. The Berquist-Sherman technique, published in 1977 CAS Proceedings and still the standard adjustment framework, separates these two effects.

The case reserve adjustment works as follows. First, compute the average case reserve per open claim at each accident year and maturity combination. Plot the average case reserve at each maturity across accident years. If you observe a systematic increase in average case reserves for recent accident years at the same maturity, this indicates changing reserve adequacy rather than (or in addition to) changing claim costs. Restate historical case incurred losses to the current reserve adequacy level by adjusting each cell so that the average case reserve per open claim matches the most recent diagonal’s adequacy standard.

The paid claims adjustment addresses the complementary distortion: changes in settlement rates. If your claims operation accelerated closures (shifting paid loss development earlier) or decelerated them (shifting paid development later, as often happens when litigation-funded claims extend timelines), the paid triangle will show distorted age-to-age factors. The Berquist-Sherman paid adjustment uses disposal rates (claims closed divided by claims reported) to detect and correct for systematic shifts in settlement speed.

After applying both adjustments, the residual uptick in your incurred development factors that cannot be explained by reserve adequacy changes or settlement rate shifts represents the true social inflation signal. This is the component that should flow into your LDF selection.

Tail Factor Selection When Nuclear Verdicts Accelerate

Social inflation creates a second problem beyond the medium-tail maturities: it distorts tail factor selection beyond the 10th development year. Standard tail fitting methods, such as fitting an exponential decay curve to historical link ratios at later maturities, assume that the rate of development diminishes smoothly toward zero. When nuclear verdicts are accelerating at the extreme tail, this assumption breaks down.

Lockton’s December 2025 report documented 135 nuclear verdicts totaling $31.3 billion in 2024, double the 2023 figure. Thermonuclear verdicts exceeding $100 million increased 81.5% year over year, with five exceeding $1 billion. These outcomes occur at late development periods, often beyond the 10th year for occurrence-basis GL policies, and they distort the tail in ways that historical curve fitting will miss.

The practical adjustment is to fit the exponential decay curve to the most recent three to five diagonals rather than to the all-year average at each maturity. If the recent-diagonal curve produces a higher tail factor than the all-year fit, use the higher value. For books with meaningful excess casualty exposure, consider a scenario-based tail where you weight the exponential fit at 50% to 70% and a Pareto-based fit (which has a heavier tail) at 30% to 50%, reflecting the possibility that extreme verdict frequency continues to accelerate.

Excess Loss Factor Recalibration

For umbrella and excess casualty pricing, the excess loss factor (ELF) is the lever where social inflation hits hardest. The ELF represents the proportion of total losses expected to pierce a given attachment point, and it is directly sensitive to the shape of the severity distribution above $1 million.

The CAS/III report provides the data foundation for recalibration. Other liability occurrence severity grew at a compound annual rate of 6.8% from 2015 to 2024, more than double the 3.2% CPI rate over the same period. Commercial auto liability severity rose 78% from 2014 to 2023, a 6.6% compound rate versus 2.8% for CPI. Guy Carpenter data shows the median size of the top 100 verdicts growing at a 10.5% annual compound rate since 2014.

To recalibrate the ELF, fit a mixed exponential/Pareto severity distribution to industry large-loss data above $1 million using the CAS/III Schedule P data or your own book’s large-loss experience. Critically, incorporate verdict severity inflation as a separate trend parameter distinct from economic inflation. If economic severity is trending at 3.5% and social inflation adds another 4% to 5% at the $1 million-plus layer, the combined severity trend for excess pricing should be 7.5% to 8.5%, not the 3.5% that an economic-only trend would produce.

At common attachment points, the impact compounds rapidly. A severity distribution trended at 8% annually rather than 4% will produce ELFs at the $2 million attachment that are 15% to 25% higher over a three-year projection, and ELFs at the $5 million attachment that are 25% to 40% higher. Underpricing at these layers is where carriers have been caught most aggressively in the current reserve cycle.

Separating Economic From Social Inflation in Trend Selection

The CAS/III report’s central methodological contribution is the decomposition of total claims inflation into an economic component (driven by CPI, ECI, and medical cost indices) and a superimposed social component. This decomposition belongs in your severity trend selection.

The approach begins with fitting an economic-only severity trend using CPI or a blended economic index (CPI for general damages, ECI for wage-related damages, medical CPI for bodily injury medical costs). Then compare the economic-only projected development to actual development on your triangle. The residual, the amount by which actual severity growth exceeds the economic-only projection, is your estimated social inflation component.

For GL occurrence, the CAS/III data suggests the economic component runs around 3.0% to 3.5% annually while the social component adds 3.0% to 4.0%, producing total severity trend of 6.0% to 7.5%. For commercial auto liability, the split is similar: economic trend near 3.0% and social trend near 3.5% to 4.0%. These two-factor trend selections should replace single-factor severity trends in your rate indication.

The advantage of the decomposition is that it allows you to adjust each component independently. If economic inflation moderates (as it has in 2025 and 2026), you can lower the economic component without assuming that social inflation will follow. Patterns we have observed across multiple reserving cycles suggest that social inflation is largely independent of the macroeconomic cycle, driven instead by litigation funding availability, juror attitudes, and plaintiff attorney tactics that show no signs of abating.

Credibility Assignment on Recent Diagonals

Standard LDF selection methods assign credibility across calendar periods using volume-weighted or simple averages that give equal (or declining) weight to older observations. When the most recent diagonals contain structurally different development patterns, this approach dilutes the signal.

The adjustment is to shift credibility toward the most recent three calendar-year diagonals in your weighted-average LDF selection at the 36-60 month maturities. A practical rule: if the most recent three-diagonal average at a given maturity exceeds the all-year weighted average by more than one standard deviation of the historical link ratios at that maturity, assign 60% to 70% weight to the recent diagonals and 30% to 40% to the all-year average. If it exceeds by more than two standard deviations, assign 80% or more to the recent diagonals.

This is not curve fitting to noise. It is a recognition that the data-generating process has changed. The Insurance Thought Leadership analysis identified backward-looking benchmarks as a root cause of reserve mispricing, noting that reliance on historical loss ratios that predated the current social inflation environment led directly to the adverse development now flowing through carrier results.

Worked Example: A 2-Point LDF Increase Through the Rate Indication

Consider a mid-size GL occurrence book with $100 million in on-level earned premium. Incurred losses at 36 months for the latest reviewed accident year total $45 million. The pricing actuary must select a 36-to-ultimate LDF and project ultimate losses.

Component	Historical LDF	Adjusted LDF
Selected 36-to-ultimate LDF	1.500	1.520
Projected ultimate losses	$67.50M	$68.40M
Loss trend to rate effective date (2 yrs at 3%)	1.061	1.061
ULAE factor	1.08	1.08
Projected loss + ULAE (trended)	$77.33M	$78.36M
Loss + ULAE ratio	77.3%	78.4%
Permissible loss + ULAE ratio	65.0%	65.0%
Indicated rate change	+18.9%	+20.6%

The 0.02 increase in the 36-to-ultimate LDF produces a 1.7-point increase in the indicated rate change: from +18.9% to +20.6%. On $100 million of premium, that 1.7-point difference represents $1.7 million in annual premium need. Across a carrier’s entire GL and excess casualty book, a systematic understatement of this magnitude compounds across multiple accident years.

This is why CNA booked $106 million in unfavorable development in a single quarter. It is why AM Best found a $9 billion industry reserve deficiency. The 0.02 looks small in the triangle. It is not small in the rate indication.

Why This Matters for Pricing Actuaries

The casualty pricing environment in 2026 demands more than reflexive rate increases. It requires a deliberate re-examination of the development assumptions that underpin every GL, umbrella, and excess casualty rate indication. The specific adjustments outlined above, including Berquist-Sherman case reserve correction, recent-diagonal credibility weighting, two-factor severity trend decomposition, and ELF recalibration at excess layers, are not optional refinements. They are prerequisites for producing rate indications that reflect the losses the industry is actually experiencing.

The carriers that recognized this earliest, reflected in hard-market rate increases of 155% cumulatively since 2015 in the excess casualty market according to CRC Group data, are now positioned with rate adequacy buffers. The carriers that relied on unadjusted historical LDFs are the ones announcing reserve strengthening on quarterly earnings calls. The difference between those two outcomes starts in the loss development triangle.

Sources