How Workers Compensation Frequency Trends Differ from Severity Trends

Workers compensation is the commercial line where conflating frequency and severity trends is most likely to produce wrong pricing. The two components of loss cost trend independently: claim frequency — the number of claims per unit of payroll exposure — responds to labor market conditions, workplace safety programs, and industry mix. Claim severity — the average cost per claim — responds to medical inflation, indemnity duration, and jurisdiction-specific benefit structures. In a given accident year, frequency may be declining while severity is increasing, and the net effect on loss costs requires separate analysis of both components.

This is not primarily an academic distinction. Carriers that apply a single combined trend factor to workers comp loss costs without decomposing frequency and severity will systematically misprice the line in periods when the two components are moving in opposite directions — which describes most of the post-2015 period for commercial workers compensation.

The Post-2015 Divergence in Workers Comp Trends

Following the extended hard market of the early 2010s, workers compensation frequency declined substantially through the mid-2010s across most states. The improvement was driven by a combination of factors: tightening OSHA enforcement, employer investment in safety programs driven by experience modification pressure, and a shift in the employment mix toward office and service sector jobs with lower inherent injury frequency than manufacturing and construction.

Severity, over the same period, increased at a rate above medical consumer price index. Medical treatment cost inflation, prescription drug costs (particularly opioids, before the regulatory response to the opioid crisis), and escalating litigation costs in states with aggressive plaintiff bars drove per-claim costs upward even as the number of claims declined. The net effect was a loss cost that was roughly flat or modestly declining in nominal terms — but flat for different reasons than it appeared.

Carriers that recognized the decomposition — declining frequency offsetting rising severity — were better positioned when frequency stopped declining in 2019 and 2020. At that point, the natural frequency cushion that had suppressed loss costs for five years was no longer available, and the severity trend that had been running at 4 to 5% annually for the same period was fully exposed in the loss cost outcome. Carriers that had been treating "flat loss costs" as evidence of a well-priced line found their combined ratios deteriorating as frequency stabilized at the same time severity continued upward.

Jurisdiction-Specific Severity Drivers

Workers compensation is a state-administered program, and severity trends vary substantially by state. The variation is driven by differences in benefit schedules (permanent total disability rates, temporary total disability maximums, and medical fee schedules are all state-specific), by the rigor of medical management requirements, and by the litigation environment for disputes about disability ratings and benefit entitlement.

California historically produces above-average claim severity compared to national benchmarks, driven by its fee schedule structure, the prevalence of disputed claims in certain industries, and the cost of living adjustments built into its indemnity benefit formula. Texas, as a nonsubscription state where some employers opt out of the standard workers compensation system, presents a different actuarial picture than most other states. Florida's workers comp reforms in the mid-2000s produced a period of favorable severity trends that have since partially reversed as the legislative changes aged.

For a carrier writing workers comp across multiple states, applying a single national severity trend to all states produces geographic cross-subsidization: favorable states are effectively subsidizing unfavorable states through a uniform trend assumption. State-specific trend analysis, supported by state-filed loss cost data from the NCCI or independent state rating bureaus, is the appropriate approach for any carrier with meaningful geographic spread in its workers comp book.

Medical Severity Versus Indemnity Severity

Within severity, the distinction between medical and indemnity components is actuarially significant. Medical severity responds to healthcare cost inflation, treatment practice patterns, and fee schedule adequacy. Indemnity severity responds to wage trends, disability duration, return-to-work program effectiveness, and the state benefit structure. The two components do not trend at the same rate.

Historically, medical severity has grown faster than indemnity severity for workers compensation, driven by medical cost inflation that has consistently exceeded general CPI. The medical proportion of total workers comp costs grew from approximately 40% in the 1980s to over 60% in recent years for most lines of business, though this varies by industry — construction and manufacturing, with higher rates of traumatic injury, tend to have lower medical proportions than office and service sectors where musculoskeletal conditions with extended treatment timelines are more common.

Carriers that use a single trend factor for total severity without separately trending medical and indemnity will produce loss cost projections that are inconsistent with the components they are projecting. If the medical proportion is growing as a share of total cost, applying the same trend factor to both components will gradually understate the total projected loss cost as the mix shifts toward higher-trending medical costs.

Experience Modification and Its Interaction with Trends

The NCCI experience modification factor (e-mod) is designed to adjust individual employer premiums based on the employer's actual loss experience relative to expected. The e-mod calculation uses the employer's actual losses compared to expected losses for the employer's size and class, with a credibility-weighting that gives more weight to actual experience as the employer's payroll grows.

The interaction between e-mod and frequency/severity trends is consequential for pricing accuracy. If class-level expected losses are updated less frequently than actual experience trends, the e-mod comparison becomes stale — employers are being compared to an expected loss benchmark that no longer accurately reflects the current loss environment for their class. This is particularly relevant when frequency is declining rapidly: the expected loss per unit of payroll in the e-mod formula may lag behind the improved frequency experience, causing e-mods to credit employers too slowly for genuine safety improvements.

NCCI updates its expected loss rates annually as part of the loss cost filing process, but the update reflects historical data with a development lag. For classes where frequency is changing rapidly, the lag between actual experience and the NCCI's expected loss update can be 18 to 24 months — long enough to produce meaningful e-mod pricing errors on both ends of the distribution.

Building a Frequency-Severity Model for Workers Comp Pricing

A workers comp pricing model that explicitly decomposes frequency and severity provides more stable loss cost projections than a combined trend approach, at the cost of requiring separate data development for both components. The data requirements are straightforward: claims per unit of payroll (exposure-based frequency) and average cost per claim (exposure-based severity) by class code and accident year, developed to a consistent maturity age.

The frequency trend should be measured against the pure premium base — claims per unit of developed payroll — rather than claims per policy, because payroll growth can make frequency appear stable even when the underlying rate of workplace injury is declining. Similarly, severity should be measured on closed claims or on IBNR-adjusted open claims at a consistent maturity age, to avoid the distortion of comparing immature accident years (where many claims are still open) against mature years (where most claims have settled).

The resulting frequency and severity trends can be combined multiplicatively to produce a pure premium trend factor, which is then applied to the historical pure premium data to project future loss costs. The advantage of the decomposed approach is that it preserves the ability to explain the projected loss cost in terms of observable components — if the model projects a 6% loss cost increase, the explanation can be "3% severity trend driven by medical inflation, offset by -2% frequency improvement from safety programs, net 6% on the medical proportion" rather than a single opaque combined trend factor.

Conclusion

Workers compensation loss cost trends should be decomposed before they are applied. Frequency and severity have different drivers and respond to different market conditions — treating them as a single combined trend obscures the source of loss cost movement and produces pricing errors when the two components diverge, which they do regularly. The actuarial work required for decomposition is standard practice in any serious workers comp pricing analysis; the benefit is pricing accuracy across market cycle phases, not just in the periods when frequency and severity happen to move together.