NCCI: Payroll Overstates Risk for High-Wage Workers, Straining Class ELRs

NCCI's July 2, 2026 research brief found that workers compensation loss ratios decline systematically as injured workers' wages rise, driven primarily by falling claim frequency, while indemnity severity plateaus above 150% of the state average weekly wage (SAWW) because statutory benefit caps limit payment regardless of actual earnings. Only 7% of injured-worker wages exceed that threshold (NCCI, July 2026), but the concentration of high earners in specific class codes makes the gap a pricing problem, not a footnote.

What NCCI Found in the Wage-Tier Data

NCCI's Insights team broke claim frequency and severity into 25-percentage-point wage tiers relative to each state's average weekly wage and tracked how both moved as wages climbed (NCCI, "The Relationship Between Benefits and Wages," July 2, 2026). Claim frequency fell as wages rose across nearly every industry group in the study, with Leisure & Hospitality and Transportation & Warehousing as the notable exceptions (Risk & Insurance, July 2026). NCCI's prior tenure research supplies the mechanism: short-tenured workers are "close to twice as likely to suffer work injuries than full-tenured workers," and higher wages within a class code correlate with longer tenure, supervisory duties, and less physically demanding tasks, each of which independently reduces claim probability per payroll dollar.

Severity moved differently. Combined indemnity and medical severity rose roughly 15% per 25-point wage tier on average, but the composition of that growth shifted as wages climbed. Indemnity severity grew about 21% per tier at the lower end, driven by the direct wage-replacement formula in most states (typically two-thirds of pre-injury average weekly wage), then flattened once workers crossed 150% of SAWW. Medical severity kept growing at roughly 10% per tier across essentially the full range, because medical treatment costs are not subject to a wage-linked statutory ceiling the way indemnity benefits are (NCCI, July 2026). The 150% threshold is not arbitrary: in a state paying two-thirds compensation subject to a maximum weekly benefit equal to 100% of SAWW, indemnity benefits cap out exactly at 100% divided by 67%, or 150% of SAWW. Above that point, a worker's actual pre-injury wage keeps rising, but the check they receive after a disabling injury does not.

NCCI's own framing draws a careful line: "payroll continues to function effectively as an exposure base because of its practicality, transparency, and consistent application across a wide range of employers," but "payroll's relationship to aggregate loss experience is not consistently uniform across all wage levels" (NCCI, July 2026). The organization is not proposing to abandon payroll. It is documenting, in its own research, a structural crack that pricing actuaries now have to reckon with directly.

How the Crack Enters the ELR

A class code's Expected Loss Rate is the anchor of manual premium: expected losses divided by expected payroll, expressed as a rate per $100 of payroll, derived from several years of the class's own experience blended with broader industry group experience using NCCI's credibility formula, then developed to ultimate, trended to the prospective policy period, and adjusted for intervening benefit changes (NCCI, "Understanding Loss Cost Actions"). The derivation assumes proportionality: a class's expected losses move in lockstep with its payroll, so that if payroll doubles, whether from adding headcount or raising pay, expected losses double too. That assumption holds reasonably well across the bulk of the wage distribution, where NCCI's own data sits, and breaks down precisely where the wage-tier research says it breaks down.

Take a hypothetical class code with 1,000 workers averaging 120% of SAWW and a current loss cost of $1.40 per $100 of payroll. Over three years, competitive wage pressure in that occupation pushes the average up to 180% of SAWW, with no change in the underlying job duties, safety program, or claim frequency per worker. Under unlimited payroll, the class's premium base grows in direct proportion to the wage increase, roughly 50% larger in dollar terms. But the loss side has not grown 50%. Frequency per worker is flat or, per NCCI's tiered data, likely lower at the new average wage tier. Indemnity severity growth slows sharply once the workforce average crosses 150% of SAWW, because the marginal dollar of wage above that point adds zero marginal indemnity benefit. Medical severity keeps climbing, but medical alone does not carry the full 50% payroll growth on its own. The ELR calculated from that class's historical experience, now measured against a payroll base that outran expected losses, is overstating the risk of every employer in the class whose workforce sits above the wage tier where the plateau begins.

The distortion compounds for individual employers whose workforce mix skews toward the top of a class code's wage range even when the class average does not. A professional-services firm paying senior staff well above the class average absorbs an ELR calibrated to the full class, including lower-wage workers whose frequency and severity patterns pull the blended rate up relative to what the firm's own high-wage-skewed workforce would generate. That employer's manual premium, before any experience modification is applied, already embeds the overstatement NCCI's research describes.

Why Experience Rating Only Partly Fixes It

NCCI's Experience Rating Plan exists to adjust an individual employer's premium away from the class average based on that employer's own loss history, and in principle it should correct exactly this kind of systematic mismatch: an employer whose true expected losses run below its class ELR will, over enough policy periods, accumulate better-than-expected actual losses and earn a credit modification factor. In practice the correction is imprecise because the experience rating formula's expected losses are themselves derived from the same class ELR that embeds the wage-tier overstatement. The plan splits each claim into primary and excess portions using state-specific split points, weights primary losses heavily and excess losses lightly by design, and compares actual to expected using a credibility-weighted formula, but "expected" in that formula is the overstated baseline, not a wage-tier-adjusted one. A high-wage employer with lower true expected losses will drift toward a credit mod as experience accumulates, but the mod measures a departure from an already-inflated starting point rather than pricing the employer against its actual risk from year one.

The lag matters because experience rating uses a three-year window of prior policy periods, excludes the most recent year to allow claims to mature, and applies credibility weighting that dampens the correction for smaller accounts. A high-wage employer newly entering a class, or one whose workforce wage mix shifts faster than the rating window can absorb, spends multiple renewal cycles paying premium anchored to a class average that does not describe it.

Limited Payroll as a Structural Corrective

Where experience rating corrects at the account level after the fact, limited payroll corrects at the exposure-base level before premium is calculated. The mechanism mirrors limited losses in excess-of-loss ratemaking: just as only losses below a chosen threshold enter certain credibility-weighted calculations, a limited payroll structure caps the payroll counted per worker at a defined multiple of SAWW, so that wages above the cap contribute nothing further to the premium base. California's WCIRB already applies a version of this for executive officers, partners, and LLC members: reported payroll for those roles is bounded between a minimum and maximum regardless of actual compensation, currently $66,300 to $171,600 effective September 1, 2026, prorated for partial-year officer status (WCIRB California, Executive Officers and Partners guidance). The rationale is functionally identical to what NCCI's wage-tier research describes for the broader workforce: above a certain compensation level, additional payroll dollars do not track additional expected loss dollars, so the exposure base should stop growing with them.

Washington State's public workers compensation fund goes further, using hours worked rather than payroll as the exposure base entirely, with rates set as a fixed rate per hour regardless of wage level (Washington State Department of Labor & Industries, Base Rates). That structure eliminates wage-level sensitivity outright: a $30-per-hour worker and a $90-per-hour worker in the same risk classification generate identical premium per hour, because the rating unit is employment exposure rather than compensation. It sacrifices payroll's practicality and transparency, the qualities NCCI credits for its durability as an exposure base, in exchange for a measure that tracks frequency exposure more directly than earnings level does.

Calibrating a per-worker payroll cap for NCCI-jurisdiction classes would start from the same 150% SAWW threshold the wage-tier research identifies as the point where indemnity severity flattens. A cap set near that level would count payroll dollars proportionally up to the point where they still carry proportional expected loss, then flatten the exposure base exactly where the loss relationship flattens. Set the cap too low, and the exposure base stops tracking real payroll growth for a large share of moderate earners, understating premium the way Nevada's now-superseded $36,000 universal cap did before SB 317 raised it to a wage-indexed threshold near $98,400 in October 2026 (see our prior analysis of Nevada's payroll-cap compression). Set it too high, and the cap never binds for the wage tiers where the overstatement actually occurs.

The Redistribution Problem

A payroll cap is a zero-sum move within a class code, not a net reduction in the class's collected premium. NCCI calibrates loss costs so that, at the state level, aggregate premium continues to match aggregate expected losses; capping payroll for high-wage workers pulls their premium contribution down toward their true expected loss share, and that shortfall has to be recovered somewhere, typically through a higher rate per $100 of payroll applied to every worker in the class, including those below the cap. Employers whose workforce concentrates below 150% SAWW would see their manual premium rise modestly to offset the reduction for employers concentrated above it. That is a defensible outcome if the current uncapped ELR is genuinely overstating risk for high earners and understating it for the rest of the class, which is exactly what NCCI's frequency and severity data imply, but it is a redistribution that state rate filings would need to disclose and defend, not a free efficiency gain.

The equity question sharpens for class codes with wide internal wage dispersion, such as professional, scientific, and technical service classes covering both junior staff near the state average wage and senior professionals well above 150% SAWW. A payroll cap calibrated to the class average shifts cost toward junior-heavy employers and away from senior-heavy ones, meaning two employers in an identical class code with identical total payroll could see materially different premium changes purely from workforce wage composition. A filing introducing a cap needs to show that distributional effect across a representative sample of employers, not just the class-level average impact.

The Trend-Selection Trap

The more immediate risk for pricing actuaries who never touch the exposure base at all is a trend-selection error hiding inside the loss ratio history itself. If aggregate wage levels in a class code rise over the experience period, whether from broad wage inflation or a mix shift toward higher-paid roles within the code, the observed loss ratio will improve mechanically as frequency drifts down the wage-tier curve, even if the underlying claim-generating process for any individual worker has not changed at all. An actuary reading a favorable loss ratio trend off historical data and selecting a correspondingly favorable prospective trend factor would be extrapolating a wage-composition effect as if it were a genuine improvement in workplace safety or claims management, then baking an inadequate loss cost into the next filing.

The correction is diagnostic rather than structural: separate the observed loss ratio trend into a wage-composition component and a true frequency/severity component before selecting a prospective trend, which requires tracking average wage per worker within the class alongside the loss ratio itself. NCCI's wage-tier research supplies the shape of that decomposition; applying it to a specific state's rate filing is state-specific ratemaking work the July 2026 brief does not attempt for every jurisdiction. Where a class code has experienced meaningful wage growth, a favorable observed loss ratio trend deserves that scrutiny before it drives a rate decrease.

Why This Matters

NCCI's own conclusion, that payroll remains "practical, reliable, and broadly applicable," is not a dismissal of the problem; it is a statement that the correction belongs in exposure-base design and trend selection rather than in abandoning payroll altogether. For pricing actuaries, the immediate work is threefold: audit class codes with wide internal wage dispersion for embedded overstatement, decompose historical loss ratio trends for wage-composition effects before selecting prospective factors, and evaluate whether limited payroll, calibrated to the 150% SAWW threshold the research identifies, belongs in a state filing for the classes where the distortion is largest. California's executive officer cap and Washington's hours-worked system show the mechanism already works in production; the open question is how far NCCI states extend it beyond officers and partners into ordinary high-wage class codes. That question sits alongside California's separately approved 10.4% pure premium increase for policies effective September 1, 2026, driven by cumulative trauma claim frequency and rising medical and allocated loss adjustment expense (Insurance Journal, July 13, 2026), a reminder that wage-tier distortion is one factor among several actuaries are reconciling in the same rate filings this year.

Further Reading

• Nevada's 21.6% WC Loss Cost Hike Exposes Payroll-Cap Pricing Gap
• NCCI Employment Surge Raises New-Worker Frequency Risk for WC Pricing
• NCCI Reserve Redundancy Erosion Signals Workers Comp Pricing Cycle Turn
• NCCI's 2024 Frequency-Severity Split Tests Trend Selection Discipline
• Tennessee WC Loss Costs Rise 4.8% as Medicare Conversion Factor Reset Bypasses Experience Ratemaking

Sources

• NCCI, "The Relationship Between Benefits and Wages" (July 2, 2026)
• Risk & Insurance, "Payroll May Overstate Expected Losses for High-Wage Workers, NCCI Research Finds" (July 2026)
• WorkCompWire, "NCCI Report Examines Relationship Between Benefits and Wages" (July 2026)
• NCCI, "Understanding Loss Cost Actions"
• WCIRB California, "Executive Officers and Partners"
• Insurance Journal, "California Insurance Commissioner OKs Upping Workers' Comp Pure Premium 10.4%" (July 13, 2026)
• Washington State Department of Labor & Industries, "Base Rates"