NCCI’s 2026 State of the Line, published in May 2026, contains a finding that most trade coverage has framed as encouraging: construction, the largest WC industry sector at 27% of premium, posted the steepest frequency decline of any major sector at 7% between 2023 and 2024. Fewer claims per payroll unit. Safer jobsites.
The rest of the sentence is the problem. Construction simultaneously posted the steepest severity increase of any major sector, 13%, more than three times the all-industry average of 4%. NCCI data on mega claims shows construction accounts for more than 40% of all WC claims exceeding $10 million, a share that dwarfs every other sector. The industry is generating fewer claims, but the claims that survive are catastrophic, and the development tail on a traumatic brain injury, spinal cord injury, or multilevel crush injury runs a decade or more.
Who it affects
Self-insured general contractors, public entities with road and infrastructure programs, and captives writing construction trades face the sharpest exposure. Employers with self-insured retentions between $500,000 and $2 million are in a particularly acute position: the retained layer is large enough to absorb a mega claim in full, but the employer’s own history rarely generates enough $5M-plus events to calibrate a credible internal frequency assumption. NCCI’s industry benchmark fills that gap, and right now that benchmark is running the wrong severity trend for this sector.
The reserve mechanism
Standard chain-ladder development patterns calibrated to the broad WC market embed the all-industry severity trend of 4%. For construction-class exposures, the relevant trend is 13%. Any IBNR estimate built on blended industry development factors uses a severity assumption that is three times too low for the largest WC sector.
The long tail compounds the problem. Medical expenses represent roughly 90% of total incurred losses on WC mega claims, per NCCI. Those payments, through surgeries, permanent disability ratings, and life care plan costs, do not emerge quickly. A construction claim that will ultimately settle for $8 million may show $1.5 million in incurred losses at the end of its third year. Development factors pulled from an industry-blended table will not project that claim to its true ultimate because the construction tail is materially longer than the blended factor assumes.
The severity trap is self-reinforcing: a 7% frequency decline reduces the number of data points in the loss triangle, which narrows the confidence interval and makes the reserve estimate look more stable, even as the claims that remain in the tail are individually larger and slower to develop than before. The industry-state frequency breakdown behind this divergence is documented in NCCI: WC Frequency Is Falling on Average, Rising in Pockets.
What this means for your next review
If your program includes a construction exposure class, ask your actuary whether the severity trend applied to those claims uses the sector-specific 13% figure or an all-industry benchmark. If the latter, request a sensitivity run at 13% trend with a construction-credentialed tail factor. That range, not the blended benchmark estimate, is the number that belongs on your balance sheet. The WC IBNR explainer covers the mechanics of how tail factors enter the reserve calculation.