Workers compensation net written premium fell 0.2% to $41.6 billion in 2025, with private carriers posting a calendar year combined ratio of 91%, the twelfth consecutive year of underwriting profitability (NCCI, 2026 State of the Line). Construction, the line's single largest industry segment at 27% of premium, recorded a 13% increase in lost-time medical severity in accident year 2024, against a 4% all-class benchmark for the same period. These two metrics sit in the same dataset and they are not telling the same story. System profitability does not guarantee class-level adequacy, and the construction signal forces a trend test that the aggregate combined ratio cannot answer.

91%
CY 2025 combined ratio, 12th consecutive year of WC underwriting profitability; accident year result was 102%
13%
Construction lost-time medical severity increase in accident year 2024, against 4% for all industries combined
$14B
Estimated industry reserve redundancy at year-end 2025, down from $16B a year earlier and $18B two years earlier

What the Calendar Year Result Conceals

The 91% calendar year combined ratio is not the metric that tells actuaries whether current rate levels will hold. The accident year 2025 combined ratio is 102%, representing actual cost experience on policies written in 2025 before prior-year reserve adjustments improve the headline number (NCCI, 2026 State of the Line). The 11-point gap between the two measures reflects favorable development on older accident years flowing through the calendar year income statement. Estimated industry reserve redundancy stood at $14 billion at year-end 2025, down from $16 billion a year earlier and $18 billion the year before that. The cushion that drove sub-90% calendar year ratios for most of the decade is contracting at roughly $2 billion per year.

NCCI Chief Actuary Donna Glenn framed the aggregate result precisely at the 2026 Annual Issues Symposium: "Behind this year's combined ratio of 91, factors such as industry mix, state differences, and carrier variation are all shaping results." That caveat is the pricing actuary's entry point. The system average describes the result; the pricing question is whether each component of that average is individually adequate. California, accounting for roughly 20% of industry premium, ran an accident year 2025 combined ratio of 129, a data point that illustrates how dramatically loss experience can diverge from the national average at the sub-aggregate level (NCCI, 2026 State of the Line). Class codes within a state can diverge by similar or greater magnitudes.

The same arithmetic applies to industry groups. A single combined ratio derived from 27% construction exposure, 20% services, and the remaining distribution of industry codes will mask divergence within each segment. The 13% construction medical severity figure did not materialize in isolation; it influenced the all-industry 4% average precisely because construction is large enough to pull the composite.

Frequency Deceleration in 2025

Lost-time claim frequency declined 2% in 2025, a favorable result on its own but a material deceleration from prior periods (NCCI, 2026 State of the Line). The long-term average annual decline runs approximately 3.8% per year across the NCCI countrywide dataset. The prior year had seen a 5% decline, ahead of that average. The 2025 deceleration to 2% is the first time in several years that frequency improvement has fallen below the long-term trend rate.

For construction specifically, NCCI reports that lost-time frequency declined nearly 7% between accident years 2023 and 2024, the largest rate of decline among all industries, and cumulative construction frequency has fallen approximately 40% since 2015 (NCCI, 2026 State of the Line). That safety record is real, and it is the mathematical source of the severity problem. As lower-severity injuries become increasingly preventable through improved fall protection, ergonomic handling equipment, and site safety programs, the remaining claims population tilts toward the injuries that cannot be engineered away: complex musculoskeletal injuries, spinal surgical cases, crush injuries, and high-fall trauma. The frequency decline selectively removes the cheap claims; it leaves behind a pool where average cost per claim rises even if no individual injury type is getting more expensive.

Medical Severity: Utilization Outpacing Price

Medical claim severity rose 4% to an average of $30,600 per lost-time claim in 2025, with indemnity severity rising 4% in parallel to $31,300 (NCCI, 2026 State of the Line). The medical severity increase outpaced the Workers' Compensation Weighted Medical Price Index, a fee-schedule-weighted measure of medical price inflation. The gap between the WCWMI and observed severity indicates that utilization growth, specifically increased inpatient and facility service use, is now the larger driver of cost increases, not pure price inflation.

This distinction shapes both the trend selection and the carrier's ability to offset cost growth through fee schedule management. A state workers compensation medical fee schedule constrains unit costs per procedure; it has limited effect on how many procedures a claim accumulates. Inpatient utilization is particularly difficult to constrain because hospital per diem or case-rate structures allow cost growth through length-of-stay extension or case complexity reclassification even when individual procedure rates are regulated. If the severity acceleration visible in both the all-class average and the construction outlier is primarily utilization-driven, then selecting a trend anchored to fee schedule inflation will persistently understate prospective severity.

The Construction 13%: Magnitude and Mechanism

The accident year 2024 increase in lost-time medical severity was driven predominantly by the construction industry, due to both the magnitude of the change and the industry's weight in the total book (NCCI, 2026 State of the Line). A 13% severity increase in a segment representing 27% of premium does not merely present a class-level pricing problem; it contributed materially to moving the all-industry composite from whatever its underlying non-construction trend would have been toward the observed 4% figure.

Three candidate mechanisms warrant evaluation in the trend analysis, though NCCI's published data does not disaggregate the 13% by driver. Injury mix is the most structurally persistent. Construction claims concentrate in musculoskeletal injuries, spinal conditions, and traumatic extremity injuries, all carrying above-average costs for surgical intervention, implants, and extended physical rehabilitation. The frequency decline described above removes lower-severity claims from the base, leaving a pool with a higher share of complex, high-cost cases. This is a mix-shift effect that appears as a severity trend even in a stable-cost environment; it is not correctable through fee schedule tightening because the underlying injury complexity, not the unit cost of any single service, is what drives the increase.

Utilization intensity is the second candidate, consistent with the system-wide finding that inpatient and specialty services are rising faster than fee schedule prices. Construction workers are, on average, older than in some other industries and carry higher rates of pre-existing comorbidities including hypertension, diabetes, and prior musculoskeletal injury that complicate treatment and extend recovery time. A 55-year-old ironworker with a lumbar disc herniation and pre-existing spinal degeneration will consume materially different medical resources than the average claim in the all-class distribution. Aging workforce comorbidity effects in workers compensation have been documented in NCCI research and amplify the utilization trend beyond what demographic-neutral pricing captures.

Reporting and development distortion is the third possibility. If construction claims from accident year 2024 are still actively developing as of the NCCI measurement date, part of the 13% figure could reflect favorable reserve releases on earlier accident years depressing those years' ultimate severity, making the 2024 year-over-year comparison appear larger than the underlying cost trend. This distinction matters because development-driven apparent severity increases should be addressed through loss development factor selection, not prospective trend loading.

Credibility Weighting a Class-Level Severity Load

NCCI's class ratemaking for workers compensation distributes a statewide indicated change across individual classification codes using credibility weighting, where the class's own experience receives a weight proportional to its statistical reliability (NCCI, Class Ratemaking for Workers Compensation). Construction's 27% aggregate share of premium implies the industry group is large enough to carry meaningful credibility in a countrywide analysis, but individual class codes within construction, such as carpentry, roofing, structural steel erection, and heavy equipment operation, each have separate experience pools that may carry credibility weights of 0.2 to 0.6 depending on claim volume in a given state.

For a pricing actuary selecting a medical severity trend for a workers compensation loss cost filing, the starting point is the statewide trend derived from NCCI's countrywide regression, credibility-blended with state-specific data. That statewide medical severity trend currently sits near 4% for most NCCI jurisdictions, reflecting the all-class figure from the 2025 experience. The practical question is whether the construction-specific 13% from accident year 2024 warrants a load above that statewide selection, and if so, by how much.

A single year observation of 13%, however large, does not fully displace a longer-fitted trend. NCCI fits exponential regression over multiple window lengths; one data point above the fitted line increases the slope estimate but may not shift a ten- or fifteen-year regression materially. The analytical test is whether the 13% movement represents a permanent level change, a genuine acceleration in the trend slope, or a transient single-year spike. A level change propagates into prospective loss costs through the trend projection; a transient spike does not, and loading the full 13% into the indication would produce a premium increase that reverses in subsequent years as the spike falls out of the experience window.

Credibility theory provides the operational framework. If the construction industry's experience in a given state carries a credibility of Z against the statewide benchmark, the class-indicated medical severity trend selection would be:

Selected Trend = Z × 13% + (1 − Z) × 4%

At Z = 0.40, the selection is 7.6%. At Z = 0.30, it is 6.7%. A reasonable range for the construction medical severity load, after accounting for credibility and before investigating the development vs. utilization question, falls between 6% and 8%, against a 4% statewide selection for non-construction classes. For a class code where medical losses represent half of total losses and the prospective trend period runs two years, the difference between a 4% and a 7% medical severity selection generates a cumulative loss cost differential of roughly 6 percentage points, enough to shift a filing from a moderate decrease to flat or from flat to a small increase.

Filing Support Under Competitive Pressure

Written bureau premium in NCCI states is expected to decrease by an average of 5% from 2025 to 2026 based on approved NCCI filings, reflecting a market environment where twelve consecutive years of profitability have driven loss costs down across most jurisdictions (NCCI, 2026 State of the Line). Proposing a construction class medical severity load in the 6-8% range, against that backdrop, requires documentation that separates class-level adequacy from the system-level profitability narrative that regulators and carriers are accustomed to citing.

The documentation structure is straightforward but must be made explicit in the actuarial report. The calendar year combined ratio of 91% reflects prior-year reserve development on a redundancy pool that has contracted by $4 billion over two years. The accident year combined ratio is 102%. The construction-specific medical severity trend is three times the all-class figure, driven by an industry that represents more than a quarter of total premium. A cross-subsidy exists in current rate levels where classes with favorable loss experience are implicitly supporting a construction severity trajectory not yet fully recognized in prospective costs. That cross-subsidy is sustainable while reserve redundancy remains; the $14 billion balance and its direction suggest the window is not indefinite.

The 2026 State of the Line data does not compel an immediate construction surcharge in every filing jurisdiction. It does compel a trend selection process that tests the statewide severity assumption against each industry group, rather than applying a single all-class trend to a book with documented internal divergence. Carriers that set construction loss costs on the 4% all-class trend today, without testing the class-level signal, are making a bet that the construction severity acceleration is transient. That may prove correct. But the 2026 SOTL data provides the actuarial basis for a more conservative position, and the shrinking reserve buffer reduces the cost of being wrong if it does not.

Further Reading

Sources

Actuarial Pricing Loss Reserving

Stay ahead with daily actuarial intelligence - news, analysis, and career insights delivered free.

Subscribe to Actuary Brew Browse All Insights