Experience modification rate (EMR): how injury history sets workers compensation premiums in the United States

Below 1.00 is a credit and lowers the premium. Above 1.00 is a debit and raises it. The number is produced from payroll and loss data by a rating organization, and it is the clearest place in American business where an employer’s injury record is converted directly into cash.

What is an experience modification rate?

The EMR is a multiplier, not a rate in itself, which is why the rating organizations call it an experience rating modification, or mod. It is applied to manual premium after classification rates have been set. NCCI’s guide shows the arithmetic plainly: a $100,000 premium at a mod of 0.75 becomes $75,000, and at 1.25 becomes $125,000. That is worth holding on to when a bid form asks for a “1.0 EMR”: it is setting a threshold on a multiplier, not on a price.

Manual rating groups every employer into a classification and charges the class average. Experience rating refines that “by comparing the experience of individual employers with the average employer in the same classification”.

A mod of exactly 1.00 does not always mean average performance. NCCI applies a unity factor of 1.00 where an employer fails the eligibility or minimum data requirements, is a new business with no data, or could not supply data after an ownership change. A new contractor showing 1.00 on a prequalification form may simply have no history at all.

Nor is it optional: experience rating “is a mandatory plan that applies to all employers that meet a state’s premium eligibility criteria”.

Who calculates an EMR, and does NCCI do it everywhere?

No. The employer’s insurer does not calculate the mod. The insurer files payroll and loss data on unit statistical reports, and a rating organization calculates the factor from it (NCCI).

NCCI is the largest of those organizations, and its Experience Rating Plan is approved in 39 jurisdictions. It is not national. NCCI states that the Plan “does not apply in California, Delaware, Michigan, New Jersey, New York, or Pennsylvania”, nor in the four monopolistic states of North Dakota, Ohio, Washington and Wyoming, which run their own plans and rates.

The Plan applies in Indiana, Massachusetts and North Carolina, but the independent bureaus there produce their own intrastate mods, and Minnesota and Wisconsin take part only where an employer has exposure in two or more participating states. The Indiana Compensation Rating Bureau lists the independent bureau states as “California, Delaware, Indiana, Massachusetts, Michigan, Minnesota, New Jersey, New York, North Carolina, Pennsylvania, and Wisconsin” (ICRB).

California is the largest exception and works to its own rules. The Workers’ Compensation Insurance Rating Bureau of California administers the California Workers’ Compensation Experience Rating Plan, 1995, and files amendments with the state Insurance Commissioner. Its September 2026 filing sets an eligibility threshold of $11,700 in expected losses, raised from $10,800, and publishes discount ratios against a range of primary thresholds rather than one fixed split point (WCIRB filing REG-2026-00001, hosted by the California Department of Insurance). Those figures are taken from the bureau’s filed proposal, so anyone relying on them should confirm that the Commissioner approved them as filed.

An employer operating across state lines can therefore hold more than one mod at once, and the same loss record can produce different numbers.

How many years count toward an EMR?

Usually three, and not the three you might expect. NCCI describes the experience period as “generally based on three years” of payroll and loss data, while noting that it “could range from less than 12 months up to 45 months” (NCCI). The data comes from policies whose effective dates fall between 21 and 57 months before the rating effective date. The policy you are currently insured under is excluded.

The reason is reporting lag, not judgment. NCCI calculates the mod 60 to 90 days before the rating effective date, and insurers need not report data on a policy until 18 months after it incepts, which gives them time to value open claims.

Apply that window to a typical annual policy and an employer renewing on 1 January 2026 is rated on policies effective 1 January 2022, 2023 and 2024, which is how the guide’s Exhibit C illustrates the rolling period. The window moves forward each year, dropping the oldest year and adding the newly valued one.

The practical consequence is that the injury an organization is dealing with today will not reach its premium for roughly two years, and will then stay in the calculation for about three. That delay is the most misunderstood feature of the number.

Why do many small claims hurt more than one large one?

Because the formula is built to measure how often things go wrong more than how badly. Each ratable loss is split at a state-approved split point. The amount below it is primary loss, which NCCI says “reflects frequency”, and the amount above it is excess loss, which “reflects severity”. Primary losses carry a greater weight in the formula and therefore a greater impact on the mod (NCCI).

NCCI states the principle directly: “the Plan gives greater weight to accident frequency than to accident severity.”

Its worked comparison, using an illustrative split point of $18,500, is that an employer with ten losses of $5,000 each and an employer with one loss of $50,000 have identical total losses, but “the 10-injury employer receives a much higher mod than the 1-injury employer”. Ten small claims are all primary; the single large one contributes $18,500 of primary loss and sends the rest to the lightly weighted excess layer.

Very large claims are damped further. A state accident limitation caps each individual loss, and the amount above the cap is non-ratable and excluded entirely. Medical-only claims are damped in the other direction: most states have approved an Experience Rating Adjustment that includes only 30% of such a claim in the calculation, which NCCI says “decreases the incentive for employers to pay medical-only claims without reporting them”.

Note that split points, accident limitations and expected loss rates are approved state by state. The $18,500 and $200,000 figures in NCCI’s guide are examples, not national values.

What is an EMR actually measuring?

This is where the number gets interesting. Because the cheap end of every claim is the part that counts most, an EMR is substantially a price on how often things go wrong rather than on how badly they ended.

That reframes it. A mod is not a verdict on the worst day an organization had. It is a verdict on the volume of ordinary, mostly low-cost events that its everyday conditions produce, priced by an actuary and charged annually. NCCI’s reasoning is that severity is largely chance while frequency is a property of the employer: “for two similar employers, the one with the higher frequency of losses will generally have higher future workers compensation costs” (NCCI).

It is worth setting that alongside the recordable rates. A TRIR or a DART rate is computed from cases an employer records about itself, so it moves with internal reporting behavior as well as with conditions, and a falling recordable rate is ambiguous evidence on its own. An experience modification rate is computed from claims filed with and valued by an insurer, which is precisely why it is less exposed to how enthusiastically injuries get written down in house. The two are not rivals. A debit mod sitting alongside a flattering TRIR is a combination that usually repays a closer look at the log.

Those frequent small events are precisely the ones a weak capture process loses: the strains reported verbally to a supervisor, the minor injuries handled informally. The Experience Rating Adjustment deliberately reduces the incentive to keep them off the books, so the honest position is that they will be counted either way.

An EMR does not measure the quality of an organization’s records. It measures the conditions those records describe, and charges for them years later. What feeds it sits alongside OSHA recordkeeping, a separate federal obligation with different thresholds, and the first report of injury, the document that opens the claim whose eventual valuation lands in the mod.

Where else is an EMR used, and can it cost a contractor a bid?

Yes, in construction especially, and the evidence for that is in statute.

Public bodies collect the number. California’s Department of Water Resources requires contractors to supply their EMR for the current year and each of the three preceding years before bidding, and a contractor whose current EMR exceeds 1.25 must provide a competent, full-time person responsible for safety on all its DWR projects at its own expense (California DWR Safety Prequalification Questionnaire). The California Department of Industrial Relations model questionnaire, used by local agencies under Public Contract Code section 20101, likewise asks bidders to list their EMR “for each of the past three premium years” (DIR).

The clearest proof that the number can shut a bidder out comes from a state that banned the practice. Virginia’s procurement code provides that “No Invitation to Bid for construction services shall condition a successful bidder’s eligibility on having a specified experience modification factor”, defining that factor as a value assigned by a rate service organization under its filed uniform experience rating plan (Code of Virginia section 2.2-4302.1). A legislature does not prohibit something that nobody was doing.

So the mod does two jobs at once. It prices insurance, and it acts as a gate on revenue, which means a debit mod can cost an organization far more than the premium difference.

What can an employer influence, and what is outside its control?

The dividing line is clean. Expected loss rates, discount ratios, split points, accident limitations and eligibility thresholds are filed and approved at state level, so an employer’s influence runs only through its own payroll and loss record (NCCI).

Within that record, the mechanism responds to three things. Frequency, because primary losses dominate. The eventual valuation of open claims, since NCCI notes that employer involvement “can reduce the severity of losses once they have occurred” and the Plan carries an incentive to return injured employees to work as soon as reasonably possible. And data accuracy, including ownership: a change must be notified to the insurer in writing within 90 days, NCCI may revise the current mod and up to two preceding ones as a result, and businesses under more than 50% common majority ownership are combined into a single mod. Group structure is therefore a rating fact, not just a legal one.

One consequence of the three-year window deserves stating plainly. An organization can be running a good program today and still be carrying a debit mod produced by claims that are years old, which is not a contradiction and not evidence that the mod is broken. It is the lag doing what it was designed to do.

Nothing here is advice on how to change a mod. Any employer wanting to know its own position should read its experience rating worksheet with its broker and its rating organization.

Related terms

• TRIR, the Total Recordable Incident Rate: the recordkeeping-side frequency measure, and why it is more exposed to reporting behavior than a mod is.
• DART rate: the severity-weighted recordable measure, covering cases with days away, restricted work or job transfer.
• First report of injury: the document that opens the claim whose valuation eventually reaches the mod.
• OSHA recordkeeping: the separate federal duty that runs alongside the insurance record.

Frequently asked questions

Sources

• NCCI, ABCs of Experience Rating
• Indiana Compensation Rating Bureau, Independent Bureaus, NCCI and WCIO
• WCIRB September 1, 2026 Regulatory Filing (REG-2026-00001), hosted by the California Department of Insurance
• Code of Virginia, section 2.2-4302.1, Process for competitive sealed bidding
• California Department of Water Resources, Safety Prequalification Questionnaire
• California Department of Industrial Relations, Labor Commissioner’s Model Prequalification Questionnaire

Last reviewed: 16 September 2026

About Logincident. Logincident is a data and software company whose configurable platform captures structured evidence at the point an event happens and presents it in dashboards and reports. We are not a law firm or a claims handler, and nothing on this page is legal advice.