The Bureau of Labor Statistics (BLS) released the September 2026 Employment Cost Index (ECI) on October 30: private industry wages and salaries rose 3.1% over twelve months, while employer health benefit costs rose 6.4%. Total private compensation rose 3.4% and total benefits 3.9%. State and local government wages rose 3.2%, benefits 3.8%, total compensation 3.5%.
We covered the first-quarter read (health 5.7%, wages 3.4%, a 2.3-point gap) and the second-quarter preview. The gap has now reached 3.3 points. This piece does something different: it sizes what that gap does to a reserve when the trend is applied correctly.
The mechanism: one blended trend hides a mix shift
Many expected loss ratio (ECR) selections apply a single severity trend to a combined workers’ comp (WC) loss. That is a problem when the two halves move at different speeds. NCCI reports that medical was 53% of WC losses for accident year 2025, leaving 47% for indemnity.
Illustration only, using the ECI as a rough stand-in for each half: a 6.4% medical trend and a 3.1% indemnity trend weight to about 4.85% (0.53 x 6.4 plus 0.47 x 3.1). A wage-only trend of 3.1% on the whole claim understates that by about 1.75 points a year. Over a three-year-old accident year being projected forward, that compounds to roughly five points of ultimate cost. It is a sensitivity, not a forecast.
The mix also drifts. If medical grows 3.3 points faster, its share of the total rises each year, so the blended trend you should select next year is higher than this year’s, even if neither component accelerates.
Caveats that matter
The ECI measures employer cost per hour worked. It is not claim cost. Health benefits in the ECI reflect premiums and plan contributions for active employees, while WC medical is paid under state fee schedules with no deductibles or networks of the same kind. Treat 6.4% as a pressure indicator, not a WC medical trend. The BLS ECI program page and the release confirm that, starting with December 2026 data, the index will update its employment weights and stop including workers’ compensation costs. The next release is January 29, 2027.
Who it affects
- Self-insured employers with WC programs, especially in high-wage workforces, where indemnity is sensitive to average weekly wage indexing, which lags wage growth.
- Self-funded health plan sponsors, who see 6.4% as the cost per hour of the very benefit they are accruing; see IBNR for self-funded health plans.
- Public entities: wages at 3.2% and benefits at 3.8% for state and local workers bear on payroll exposure bases for pools and on pension and other post-employment benefit (OPEB) assumptions.
Where this shows up in your reserves
Open the actuarial report’s trend selections for open accident years. Look for one severity trend line or two. Then check the payroll exposure base used for the current year ECR: if it is last year’s payroll, a wage rise of 3.1% to 3.2% is already missing from the denominator. See Workers compensation IBNR for how the pieces fit.
Our call
Expect the 53/47 split to tilt toward medical in AY2026 and AY2027 selections, but not by the full ECI gap, because fee schedules damp WC medical. Our read is that the health line is a leading indicator for self-funded plans now and for WC medical with a lag of several quarters. Reserve adequacy risk is highest where one blended trend is carried across years. Because the ECI is changing in January, move your benchmarking to a standalone wage series and a medical series before then.
What this means for your next review
Ask your actuary to show wage and medical trends separately, with the resulting blended trend by accident year and the exposure base used. Request a sensitivity at plus and minus one point on the medical trend alone.
Decision-maker checklist
- Confirm trend selections are split between wage-sensitive and medical-sensitive components.
- Update payroll exposure bases to current-year wages before the year-end study.
- Ask what replaces the ECI as your benchmark after January 2027.
- Watch the October CPI medical care services print as a cross-check.