How a Workers' Comp EMR Is Calculated

Updated September 12, 2026

You’ve seen the worksheet. Rows of numbers, a handful of claims, a payroll table, and a single figure at the bottom that decides whether you can bid the job. Nobody hands you the formula. Here’s what’s actually in it.

If you haven’t read what an EMR actually is, your workers’ comp EMR, or experience modification rate, is a multiplier on your premium built from your own claims and payroll. Start there. This article is about how the number gets built, line by line.

Your mod is based on three years, and not last year

Your mod isn’t based on this year. It’s based on three policy years, and not the three most recent ones either. The rating bureau (NCCI in most states, a state bureau in a handful) skips the year that just ended and uses the three before it.

That gap is the one-year lag. It exists because claims take time to report and reserve accurately, and payroll takes time to audit and finalize. A mod calculated in 2026 typically uses policy years 2022, 2023, and 2024, not 2025. Something you fixed last year won’t show up in your rating yet. It’s already in the pipeline. It just hasn’t reached the front of the line.

That also means your current mod is a lagging indicator of a lagging indicator. If your worksheet looks bad, you’re looking at old news. If it looks good, the same is true. A bad year you just had hasn’t arrived yet either.

Expected losses set the baseline

For each class code you carry (roofing, electrical, concrete, whatever operations your payroll is reported under), the bureau has an expected loss rate: what a typical contractor in that class, with that payroll, is expected to lose to claims in a given year. Multiply your payroll in each class by its expected loss rate, and you get your expected losses. This is the baseline your actual claims get compared against.

Get the class codes wrong, and you’re comparing your actual losses to the wrong baseline. A framing crew’s payroll coded as general carpentry instead of framing, or an electrician’s payroll sitting under a lower-risk code from years ago, changes the expected-loss side of the equation without anyone touching a claim.

Actual losses are split into primary and excess

Your actual losses come from your claims in the same three years, valued as of a specific date (more on that below). Each claim isn’t counted at full value straight through. It’s split into two pieces.

A primary portion, roughly the first slice of the claim’s cost, counts at full weight. An excess portion, everything above that slice, counts at a reduced weight.

This split is why frequency hurts more than severity. Take two contractors who each have $90,000 in claims over three years. One has a single $90,000 claim. The other has nine claims at $10,000 each. The nine small claims are each likely to fall entirely, or almost entirely, inside the primary slice, so nearly all $90,000 counts at full weight. The one large claim has most of its value pushed into the excess portion, which counts at a fraction of that. Same total dollars, very different mod impact. This is also why “we had one bad claim” is a real explanation for a bad mod, and “we have a claim every year” is a harder pattern to talk your way out of.

The dollar amount where primary ends and excess begins, the split point, is set per state under NCCI’s plan and updated with each state’s annual filing, so check the figure printed on your current worksheet rather than assuming one.

Open claims are counted at their reserve, not what’s been paid

A claim’s value in this calculation isn’t what’s been paid out yet. It’s the reserve: the carrier’s current estimate of what the claim will ultimately cost, paid and unpaid combined. If a claim is still open on the valuation date, the reserve is what goes into your mod, whether or not that estimate turns out to be right. An overstated reserve on an open claim inflates your mod for as long as it stays open and overstated. That’s exactly why old, open claims are worth checking, not just closed ones.

The comparison, in one pass

The bureau compares your actual losses (primary plus discounted excess, across three years) to your expected losses (payroll times expected loss rate, by class code), applies a weighting that keeps a single bad year from wrecking a small contractor’s mod entirely, and produces a ratio. Above 1.00, your losses ran higher than expected for a contractor your size doing your work. Below 1.00, lower.

Every input in that sentence is a number somebody typed into a system. None of them are self-correcting. The bureau calculates from what the carrier reports, and the carrier reports from what’s in its file. If nobody puts the worksheet next to the loss runs and the payroll records, errors in any of those inputs just ride through to the bottom line.

If you want to see what a specific error would be worth on your mod, the free EMR Cost & Bid Check calculator walks through it. We’re building a full worksheet verification, and you can get on the waitlist to be first in line when it opens. It’s a flat fee, no contingency, and the savings are yours.