The Relationship Between Benefits and Wages
Posted Date: June 2026
Key Insights
- Although alternatives exist, unlimited payroll remains a practical, reliable, and broadly applicable workers compensation exposure base
- Indemnity severity generally rises with wages but plateaus at higher wage levels, while medical severity increases more consistently across all levels
- Claim frequency typically declines as individual wages increase, contributing to variability in the proportionality between benefits and wages
Introduction
An exposure base is the unit of measurement to which insurance rates are applied in determining premium. According to the Casualty Actuarial Society’s publication Basic Ratemaking (Werner and Modlin, 2016), a good exposure base “should be directly proportional to expected loss, be practical, and consider any preexisting exposure base established within the industry.”
Since the 1970s, unlimited (total) payroll has served as the exposure base for nearly all NCCI classification codes and states, in part because payroll is objectively determined and readily verifiable. It satisfies the practicality and industry precedent criteria. This research brief will explore the proportionality criterion, examining the relationship between benefits and wages.
The choice of the exposure base directly impacts class loss costs and rates, which serve as industry benchmarks and are embedded in carrier pricing. More importantly, it broadly determines how premium is calculated and distributed across insureds, subject to experience rating modifications that account for each insured’s claims experience. Therefore, it’s important to periodically reassess the relationship between the exposure base and associated losses.
Background
NCCI’s last exposure base review (2006) concluded that “total payroll as an exposure base is an appropriate reflection of loss potential for all classes and for each industry group.”
In isolation, using unlimited payroll as an exposure base implies that as a company’s payroll grows, its expected losses grow proportionately. Published loss costs vary by state (s) and class (c), and the manual premium on a policy in its simplest form is expressed below.
\[ {Manual\ Premium} = \sum_{s,c}Payroll_{s,c} \times Loss\ Cost_{s,c} \]
The actual policy premium will differ due to company expenses, experience rating modifications, and various other adjustments.
Payroll doesn’t control for the number of employees or differences in salaries across similar companies. Consider two companies in the same state, in the same business, employing the same number of employees. If one company pays twice the wages of the other, its manual premium would also be twice as high. This is appropriate if losses were also twice as high, but in theory, there are several components of total losses that may not be expected to directly vary with wages. This example is directly discussed in Exposure Bases Revisited (Bouska, 1989), which suggests “to the extent that the losses arise from medical payments or are capped by the maximum benefits payable under state law, the [premium] difference is not justified in terms of expected losses.”
One notable limitation of NCCI’s prior study was the lack of injured worker wage data, as only a sample of claims was reported to NCCI at that time. With the 2020 inception of the Indemnity Data Call (IDC), we can now analyze injured worker wages on all lost-time claims. This newly available data, along with continually progressing industry research, prompted another exploration into the unlimited payroll exposure base.
Data
The primary data sources underlying this brief are the Unit Statistical Plan Data (unit data) and the IDC data. To calculate severity, frequency, and loss ratios, we used unit data exposure and incurred losses on lost-time claims at a first report, limited to $500K from Policy Years 2022 and 2023. The IDC data provides individual employee wages on lost-time claims to accurately segment claims into wage tiers. We also leveraged NCCI’s Policy Data to classify policies by industry, using the reported North American Industry Classification System (NAICS) codes.
The distribution of claims by type is shown to the right. The analysis includes lost-time claims that can be linked between unit data and IDC data, as these contain individual wage information. Med-only claims account for the majority of claim counts but represent a small share of total losses due to their relatively low severity. Roughly 20% of all claims are included in the analysis, which captures over 67% of the overall losses.
Because we do not have insight into the employee-specific wage levels underlying the majority of payroll reported to NCCI, we leveraged state and industry wage distributions from the Occupational Employment and Wage Statistics (OEWS) data, produced by the U.S. Bureau of Labor Statistics through a large, stratified employer survey. More detail on the application of this data is described in the Frequency section below.
The lack of detailed wage data also precluded evaluating the impacts of experience rating modifications.
The Relationship Between Benefits and Wages
Loss costs are the basis for calculating premium for a given class code and can be decomposed into frequency (claim occurrence) and severity (individual claim cost).
\[\text{Loss Cost}=\frac{\text{Expected Losses}}{\text{Exposure}}=\text{Frequency}\times\text{Severity}\]
Because premium is intended to be commensurate with expected losses, exposure should ideally scale proportionately with expected losses across wage levels, reflecting both frequency and severity. Expanding these terms, expected losses can be expressed in terms of frequency, severity, and exposure.
\[\text{Expected Losses}=\left(\frac{\# \text{ Claims}}{\text{Exposure}}\right)\times\left(\frac{\text{Loss Dollars}}{\# \text{ Claims}}\right)\times\text{Exposure}\]
This equation highlights that differences in expected losses could arise from either changes in the likelihood of claim occurrence or changes in individual claim costs. Examining these components individually helps identify what drives differences in expected losses by wage level.
Severities
Using the IDC data, we can analyze both indemnity and medical payments on all reported lost-time claims.
To appropriately aggregate across states, claims are binned using relativities instead of nominal dollar amounts. We used the state average weekly wage (SAWW) to normalize workers’ wages on each claim, as this is the standard basis of statutory indemnity benefits. In the charts below, observations in each wage tier represent claims from workers whose wages fall within a given percentage range of the SAWW (e.g., the 100–150% tier includes workers earning between 100% and 150% of the SAWW). The 0–25% tier was removed in all charts due to a low volume of exposure. Claim severities were also normalized based on the state’s average claim amount.
As shown in the chart to the right, total (indemnity + medical) severity generally increases as wage tier increases. This aligns with conclusions from prior publications, and, in isolation, may indicate that benefits continue to grow with wages at all levels. However, it’s notable that claim costs increase at a different rate than wages. On average, moving up one 25% wage tier corresponds to roughly a 15% increase in combined severity, though the relationship varies materially by state, industry, and class code.
The following sections examine this relationship in greater detail.
Indemnity Severity
Indemnity severity is expected to rise with wages because benefits are directly tied to an injured worker’s wages. The figure below examines indemnity severity in greater detail for all industries combined and for each individual industry. Average indemnity claim costs generally track well with wages at lower tiers, but they begin to plateau for workers whose wages are above 150% of the SAWW.
This pattern generally holds across all industries. The observed volatility in higher wage tiers can be attributed to thin data, as just 7% of injured worker wages reported to NCCI are more than 150% of the SAWW.
This tapering aligns with the intuition that statutory benefit maximums take effect and limit indemnity severity growth. However, this may not fully explain the underlying relationship.
In the graph below, states are grouped according to the wage level at which the maximum indemnity benefit is reached for the benefit type with the highest statutory maximum. For example, consider a state where indemnity benefits are paid at a rate of 67% of the injured worker’s wage, subject to a maximum benefit of 100% of the SAWW. The workers’ indemnity benefits would be capped when they earn 150% (100% / 67%) or more of the SAWW, and these observations would be grouped in the “150% of SAWW” series.
The shaded areas represent the points at which statutory maximum indemnity benefits would be expected to take effect and cause the series to flatten. If statutory maximums were the sole driver of the observed plateau in average indemnity severity, we would expect to see a clear leveling of the series and meaningful differentiation across state groupings. We can see that indemnity severity broadly increases proportionally to wages, but as expected, this relationship weakens at high wage levels where statutory maximums take effect.
While statutory maximums contribute to the relationship between indemnity severity and wages, it may not fully explain it, as the maximum associated with a single benefit type does not fully represent the complexity of the state’s overall benefit framework.
Medical Severity
Medical benefits can be analyzed similar to indemnity benefits. In the chart below, we observe that medical severity tends to grow with wages at all levels. However, the increase is smaller than for indemnity—on average, moving up one 25% wage tier corresponds to roughly a 10% increase in medical severity, compared with about a 21% change in indemnity severity. Unlike indemnity benefits, there is no direct link between medical benefits and wages, and the observed relationship likely reflects multiple underlying dynamics. Highly paid workers in some industries may be performing more specialized or complex tasks and be at risk for more severe injuries. Claimant behavior, such as propensity to file a claim or utilization of medical services, may also be a contributing factor.
Based on findings from the 2006 exposure base study, the relationship of medical benefits and wage levels was further explored in a subsequent research brief, The Performance of Total Payroll as the Exposure Base for Workers Compensation—Updated (2007).
Another hypothesis is that worker age explains the observed increase in medical costs with wages, because older workers tend to earn higher wages and medical severity is known to increase with worker age. While medical severity does rise with worker age, the chart below shows that it generally does so for all wage levels. This suggests that increases in medical severity are driven by factors independent of injured workers’ age.
Frequency
Severity can be directly analyzed using NCCI data due to the availability of weekly wage information on lost-time claims. Measuring frequency is more complex because wages are only observed for injured workers. Estimating frequency by wage tier requires allocating exposure across wage tiers, which requires an all-worker wage distribution. Using the IDC for the all-worker wage distribution may not be appropriate because the percentage of premium directly represented by wages on lost-time claims is small, and previous NCCI studies have shown that the injured worker average weekly wage is lower than the all-worker average weekly wage.
NCCI regularly reviews and publishes the Injured Worker Wage Distribution to support pricing legislation that affects indemnity benefits.
All-worker wage distributions were derived using OEWS data, and these distributions were used to allocate reported payroll across wage tiers. Payroll was converted to premium using applicable loss costs, allowing payroll exposure to be normalized across class codes. Frequency was then calculated by using reported lost-time claim counts in each wage tier.
As shown below, average claim frequency declines across wage tiers for all industries combined. Several factors may contribute to this pattern. Within an occupation, claim incidence per worker tends to be lower for more tenured workers; previous NCCI research suggests “short-tenured workers in most sectors are close to twice as likely to suffer work injuries than full-tenured workers” (2022). Because wages often correlate with tenure and injury risk declines as workers gain task-specific experience, the number of claim counts may increase less than proportionately with wages across tiers.
In addition, workers across wage levels within an industry may face broadly similar workplace hazards, limiting the extent to which claim counts rise with wages. Higher-paid workers may also transition into supervisory or less physically demanding roles, reducing their exposure to occupational risk. Holding all else equal, because premium increases with wages through payroll, stable or declining claim counts translate into lower observed frequency at higher wage levels.
Frequency can also be observed by industry. Premium allocation at the industry level is susceptible to volatility, particularly at the lower wage tiers. While the decreasing pattern generally holds across industries, the Leisure & Hospitality and Transportation & Warehousing industries show increasing frequency across some wage tiers.
Loss Ratios
Loss ratios capture the combined effects of both frequency and severity by providing an overall measure of losses relative to premium. If expected losses scaled proportionally with wages, loss ratios would remain relatively flat across wage tiers. Instead, the graph below shows that loss ratios decline as wage tiers increase. Frequency is the primary driver of these loss ratio decreases. Severity somewhat offsets this, particularly in lower wage tiers, before plateauing. The loss ratio patterns for the Leisure & Hospitality and Transportation & Warehousing industries are consistent with those from the frequency graph above.
Summary Observations
In summary, the relationship between severity and wages is generally linear, but it doesn’t appear to hold at the highest wage levels. Additionally, the increase in wages generally outpaces the increase in severity, indicating that individual claim costs may not grow as fast as wages. Frequency tends to decline as wages rise, further driving declining loss ratios across wage levels. In aggregate, employers differing primarily in employee compensation may not exhibit proportionally different expected losses. Taken together, these findings suggest that unlimited payroll may be an imperfect proxy for expected losses.
In the sections below, we discuss the relationship at a class level, mitigating effects through experience rating, and practical considerations.
Class Code Differences
The relationships shown above are based on aggregated data and could differ for individual class codes. For example, in occupations where wages more closely track differences in job duties or hazard levels, payroll may remain a strong expected loss proxy. Further, in class codes with limited wage dispersion, the sensitivity to the exposure base is inherently constrained.
Beyond class codes, the effect of wage dispersion within a class may be partially offset at the individual risk level through the experience rating modification (mod). A mod compares a risk’s historical claims to its expected losses, which are calculated by applying the class-specific expected loss rate to the risk’s exposure. To the extent that higher wages don’t translate into higher claim costs, this difference would be reflected through a lower mod, better aligning premium with losses.
To review data at a class level, we examined average injured worker wages relative to the SAWW, as well as the 10th and 90th percentiles. As noted earlier, these injured worker wages may not be representative of all workers. The two class codes with the highest interdecile range from each NCCI industry group are shown below. High average wages appear in both hazardous industries (e.g., steelmaking and mining) and less hazardous occupations (e.g., broadcasting or sales). This reflects a mix of classes where higher wages may or may not correspond to greater hazard and illustrates the challenge of selecting an exposure base appropriate for all classes.
| Class Code | Class Description | 10th Perc Wage/SAWW | Average Wage/SAWW | 90th Perc Wage/SAWW | Interdecile Range |
|---|---|---|---|---|---|
| Manufacturing | 0.44 | 0.83 | 1.33 | 0.89 | |
| 4740 | Oil Refining—Petroleum & Drivers | 0.62 | 1.75 | 2.95 | 2.33 |
| 3004 | Iron or Steel—Manufacturing—Steelmaking & Drivers | 0.61 | 1.44 | 2.45 | 1.84 |
| Contracting | 0.50 | 0.95 | 1.51 | 1.01 | |
| 6213 | Oil or Gas—Well—Specialty Tool & Equipment Leasing NOC—All Employees & Drivers | 0.68 | 1.50 | 2.55 | 1.87 |
| 5160 | Elevator Erection or Repair | 0.60 | 1.49 | 2.43 | 1.83 |
| Office & Clerical | 0.37 | 0.92 | 1.61 | 1.24 | |
| 8803 | Auditor, Accountant, or Computer System Designer or Programmer—Traveling | 0.37 | 1.20 | 2.41 | 2.04 |
| 7610 | Radio or Television Broadcasting Station—All Employees & Clerical, Drivers | 0.47 | 1.38 | 2.39 | 1.92 |
| Goods & Services | 0.28 | 0.67 | 1.12 | 0.84 | |
| 8720 | Inspection of Risks for Insurance or Valuation Purposes NOC | 0.49 | 1.20 | 2.23 | 1.74 |
| 8393 | Automobile Body Repair & Drivers | 0.50 | 1.06 | 1.87 | 1.37 |
| Miscellaneous | 0.36 | 0.93 | 1.59 | 1.23 | |
| 7422 | Aviation NOC—Other Than Helicopters—Flying Crew | 0.47 | 1.40 | 2.56 | 2.09 |
| 7425 | Aviation—Helicopters—Flying Crew | 0.79 | 1.89 | 2.83 | 2.04 |
Alternative Exposure Bases
The primary limitation of unlimited payroll is illustrated above, and some bureau states use alternative exposure bases. The California Workers’ Compensation Insurance Rating Bureau has implemented payroll caps for select class codes, limiting the inclusion of high wages in payroll determination. In Washington, the State Fund uses hours worked rather than payroll, directly tying premium to employee time-at-risk rather than compensation levels.
NCCI also has some exceptions to unlimited payroll in place. The high wages of executive officers, sole proprietors, and athletes are capped to prevent large distortions in loss costs. By statute, Nevada similarly caps the payroll used in calculating loss costs. Additionally, some class codes use different exposure bases; taxicab operators are rated using a per-driver basis generally tied to the SAWW, while domestic workers are rated based on the number of employees.
Limited Payroll
Limited payroll can improve upon unlimited payroll for class codes with substantially high wages, where additional payroll is no longer expected to proportionally increase losses. Once payroll is capped, extra compensation above the limit does not add exposure. This shifts the emphasis from how much workers earn to how many workers are employed, giving claim frequency greater influence relative to severity. In class codes where wages exceed the level at which losses continue to scale with payroll, a payroll limit may improve alignment between premium and expected losses. A payroll limit may also dampen premium volatility driven by compensation changes unrelated to underlying risk.
However, applying a payroll limit introduces tradeoffs. A payroll limit redistributes premium within a class code by reducing the share attributed to higher-wage employers and increasing the relative share for lower-wage employers. If applied too broadly or at an inappropriate level, this redistribution could introduce equity concerns. In addition, because all wages above the limit contribute the same exposure, a payroll cap may concentrate exposure near the cap and increase sensitivity to small shifts in wages around the threshold. Selecting an appropriate payroll limit and determining which class codes warrant its application requires careful review and reassessment as wage distributions evolve.
Hours Worked
Outside of payroll, hours worked is the most frequently cited alternative exposure base for workers compensation. Its primary advantage is that it more directly quantifies exposure to injury, as employees who work more hours have a greater opportunity to sustain an injury and file a claim, regardless of wage level.
The main drawback of hours worked is its weak connection to severity. Indemnity benefits are tied to wages, so an exposure base that doesn’t incorporate earnings wouldn’t scale with expected costs per claim. Adding specificity may help, such as hours worked per occupation or per job function, but premiums may still understate expected losses for higher-wage paying employers and overstate them for lower-wage paying employers. In addition, an hours-worked exposure base would depend on accurate time reporting that overcomes differences in scheduling practices, overtime, and hourly versus salaried compensation structures.
Conclusion
Unlimited payroll has served as the industry’s standard exposure base for decades, though it’s not without limitations. It reasonably captures the relationship between wages and claim severity but does not consistently reflect differences in claim frequency across wage levels. As a result, aggregate loss ratios generally decline as wages increase, driven by lower frequency and compounded by capped indemnity severity.
Wage distributions vary meaningfully across classes and even more so across individual insureds. This exploration focused on aggregate relationships, but any final impacts are dependent on employers’ wage levels, experience rating modifications, and carrier pricing.
Unlimited payroll remains the industry standard as it is readily available, verifiable, and has a long-standing historical precedent. While alternative exposure bases have conceptual appeal and are used by NCCI and other bureaus in select state and class combinations, the perceived trade-offs have maintained the status quo.
NCCI will continue to monitor industry trends and update this analysis as appropriate.
References
Bouska, Amy S. 1989. “Exposure Bases Revisited.” In Proceedings of the Casualty Actuarial Society. Vol. 76. 145. Casualty Actuarial Society. www.casact.org/sites/default/files/2021-02/pubs_proceed_proceed89_89001.pdf.
Coate, Patrick. 2022. “The Great Reshuffle and Workers Compensation Frequency.” Research Brief. National Council on Compensation Insurance (NCCI). www.ncci.com/Articles/Documents/Insights-Great-Reshuffle2022.pdf.
Moore, Natasha, and Delano Brown. 2006. “The Performance of Total Payroll as the Exposure Base for Workers Compensation.” Research Brief. National Council on Compensation Insurance (NCCI). www.ncci.com/Articles/Pages/II_Research-performance-payroll.pdf.
———. 2007. “The Performance of Total Payroll as the Exposure Base for Workers Compensation - Updated.” Research Brief. National Council on Compensation Insurance (NCCI). www.ncci.com/Articles/Pages/II_Research-performance-payrollUpdated.pdf.
National Council on Compensation Insurance (NCCI). n.d. “Injured Worker Wage Distribution.” www.ncci.com/Articles/Pages/II_Injured-Worker-Wage-Distribution.aspx.
Werner, Geoff, and Claudine Modlin. 2016. Basic Ratemaking. 5th ed. Arlington, VA: Casualty Actuarial Society. www.casact.org/sites/default/files/2021-03/5_Werner_Modlin.pdf.