Same people, same work, half the rate

Ask three employers what workers compensation costs and the answers arrive in different units. One pays 158 dollars a year. One pays about 180,000. The third does not buy it at all, he advises on it, and he prices it as a share of payroll: one account he took on was paying 16.8 percent and now pays under 9 percent, for the same people doing the same work.

Money & World asked business owners, founders, office managers and freelancers carrying any commercial cover what they pay, what happened at their last renewal, and what an insurer actually paid against what they expected. Seven people answered. Three of them described workers compensation from three positions that almost never appear in the same place: the advisory side, a warehouse floor at scale, and a solo practice small enough that the premium is a rounding error.

None of the three can explain their own number by pointing at a claim they had. That is not a complaint about insurers. It is how the calculation is specified.

The advisory view, where the number is set before anybody notices

Answer: Isaac Attia says the two things that consistently cost employers more than they realise are open reserves, which price a policy regardless of what a claim eventually settles for, and the classification code, which he says almost nobody audits.

Isaac Attia is founder and independent PEO consultant at ForwardPEO. He is the only contributor here who sees many employers rather than one, and he says so before saying anything else.

"I sit on the advisory side rather than the buying side, so I see renewals across a lot of industries and states. Two things consistently cost employers more than they realize."

The first is the one that matters most, because it is invisible on any invoice.

"Open reserves price your policy, not closed claims. Whatever reserve an insurer has set against a still-open claim sits on your loss run and travels with you into every quote for as long as it's there"

That holds, he adds, "regardless of what the claim eventually settles for." Then the case he uses to show it.

"I had an account go to market carrying a large open injury reserve that dragged down every number on the table. It was later determined to be fraudulent and we requoted the business. For that whole stretch the exposure wasn't real, but the pricing very much was."

The second is a filing decision that may have been made by somebody who has since left.

"The classification code is the number almost nobody audits. The rate attaches to the code, not the job title, so a decision made years ago quietly sets the price forever."

He puts a number on what correcting one is worth. An account he took from 16.8 percent of payroll to under 9 percent was, in his words, "same people, same work, coded correctly. Nothing about the risk changed."

That is the largest single movement anybody in this pool reports, and nothing physical produced it. Then geography, which he treats as an amplifier of both.

"Geography compounds both. On a multi-state roofer I've seen the rate differ by more than half across a state line for identical work."

The same mechanism, seen from underneath

Answer: Joe Spisak says his workers compensation premium ran about 180,000 dollars a year at 80 plus warehouse employees, that it rose 14 percent after two forklift incidents in one quarter, and that a single back injury paid out north of 40,000 dollars against an expected 12,000 and pushed his experience modification rate up the following year.

Joe Spisak is chief executive of Fulfill.com. Where Attia describes the machinery, Spisak describes being inside it, and he names the same component.

"Running a 140,000 square foot fulfillment center taught me that workers comp is the silent profit killer nobody talks about until renewal hits. Our annual premium ran around $180,000 when we were at full scale with 80+ warehouse employees, and the rate jumped 14% one year after we had two forklift incidents in the same quarter. Nothing catastrophic, but the actuaries don't care about context."

The claim he describes is the textbook case for what a reserve does, because its final cost was more than three times the early estimate.

"The real shock came when one of our team members injured their back moving a pallet. We expected the claim to cover maybe six weeks of recovery based on the doctor's initial assessment. Workers comp ended up paying out for nearly nine months because the injury kept recurring. Total payout was somewhere north of $40,000 when we thought we were looking at maybe $12,000. That single claim pushed our experience modification rate up the following year, which meant higher premiums even though we'd implemented new safety protocols and gone incident-free for 18 months."

Two things are happening in that sentence and only one of them is the injury. The other is that eighteen incident-free months did not arrive in time to matter, because the modification for the following year was already built from a period that had closed.

He reports the same drift on the general liability side, in the deductible rather than the premium.

"General liability was easier to stomach at around $22,000 annually for a company doing $10M in revenue, but I watched our deductible creep from $5,000 to $10,000 over three renewals without anyone explicitly calling it out. You have to read the renewal docs line by line because insurers bank on busy founders just signing and moving on."

And then the conclusion he draws, which is the same one Attia reaches from the other side of the desk.

"The frustrating part is how disconnected premium increases are from your actual operations. We invested heavily in safety equipment, training, and warehouse automation specifically to reduce risk. Claims went down. Premiums still went up."

The control case

Answer: Luke Smith says his Minnesota law practice pays 158 dollars a year for workers compensation, down on last year, that an audit produced no change to the premium, and that raising a liability limit on his general liability policy did not change that premium either.

Luke Smith is attorney and founder of LawSmith PLLC. His firm is a work injury practice, which is worth stating plainly: the contributor describing his own workers compensation premium here also litigates workers compensation for a living. He was asked what he pays as an employer, and he answered as one.

"I run a small law practice in Minnesota. I have a commercial general liability policy I paid a premium of $333 this year. I did have to add my landlord as a third-party beneficiary and increase the liability limit in order to comply with a property lease. I was surprised that those changes (particularly the increased liability limit) did not change the premium. I had no claims."

The workers compensation line is the one that matters for the argument, and it moves in the opposite direction to Spisak's.

"I have a workers' compensation insurance policy for which I paid a premium of $158 this year, which was lower than last year. I had an audit that resulted in no change to the premium. Nothing changed about the coverage or deductible. I had no claims."

His third policy went the other way with nothing to explain it either.

"I have a professional liability policy for which I paid a premium of $2,025, which was higher than last year. There were no changes to the coverage or deductible. I had no claims."

Three policies, no claims on any of them across the life of the practice, and three different directions of travel. His summary of that is his own and is offered as such.

"I have had all of those policies since I started my practice and have never had a claim. So those insurers are doing a great job of collecting premiums and not paying claims."

Smith is the useful case precisely because nothing happens in it. A firm with almost no payroll and no losses is the closest thing this pool has to a control, and the number still moved in three directions at once.

What the manual actually specifies

Attia's central claim is checkable, because the rules for computing a modification are published. In Minnesota, which is where Smith practises, they are published by the Minnesota Workers' Compensation Insurers Association in the Minnesota Experience Rating Plan Manual. That manual governs Minnesota and nowhere else, and it is read here because it is public and because one of these three contributors operates under it.

The manual's own instruction for reporting losses is the whole of Attia's first point. Carriers are told to fill in the sum of "Incurred (paid plus reserved) Indemnity" and the sum of "Incurred (paid plus reserved) Medical".

Paid plus reserved. The amount the insurer has set aside against a claim that has not resolved counts at its reserved value. When Attia says an open reserve prices the policy and a settlement does not, he is describing the arithmetic as written.

The manual then removes the question of fault entirely, at Rule 1-C-3-a.

"No loss shall be excluded from the experience of an employer even if the employer was not responsible for the accident that caused such loss."

The single exception attached to that rule is for claims reported as attributable to the COVID-19 pandemic with accident dates between 1 December 2019 and 30 June 2023. Nothing else about responsibility is relevant.

There is a correction route, and its shape is the last piece of Attia's account. When a loss value is revised, the manual says that submission of revised unit reports "will result in the automatic recalculation of the current and up to two preceding experience rating modifications", and that the time frame for those three "is limited to the employer's fifth most recent rating effective date". The listed circumstances include a subrogation recovery obtained against a third party and a claim found non-compensable by an official ruling denying benefits.

So a wrong number can be unwound, but only within a bounded window, and only once somebody establishes that it was wrong. Attia's fraudulent reserve is exactly that case, and his description of it, that the exposure was not real but the pricing was, is a description of the interval before the correction arrives.

What geography does, measured

The largest movement anybody in this pool describes is Attia's, and the second largest is the one he attributes to a state line. That one has been measured, every two years since 1986, by a state government.

Oregon's Department of Consumer and Business Services prices the same set of industry classifications in all 51 US jurisdictions and publishes the ranking. The method is the point: it takes 53 NCCI class codes, weights them by Oregon's own payroll distribution, and applies that fixed mix to every jurisdiction's rates. Industry composition is therefore held constant, so what is left is the price of the same work in a different place.

For rates effective in January 2024, the study puts the national median at 1.09 dollars per 100 dollars of payroll. That is the lowest median in the study's history, against a peak of 4.35 dollars in 1994. North Dakota is lowest at 0.50 dollars and Hawaii highest at 2.52, a countrywide spread of 2.02 dollars. Expressed against the median, which is how the report itself prefers to express it, the range runs from 45 percent to 231 percent, a spread of 186 percent, and the report notes that this is wider than the 145 percent it measured in 2022.

Minnesota, where Smith pays his 158 dollars, sits 17th at 1.25 dollars, or 114 percent of the median. Oregon itself is 14th lowest at 0.89 dollars.

The middle of the table is where the report is most emphatic about how little separates jurisdictions: across the middle 26, rates run from 1.34 dollars to 0.80, and the average difference between one rank and the next is two cents. Small movements produce large rank changes, which is why the report tells readers to look at a jurisdiction's position as a percentage of the median rather than its rank.

And then, in its own notes about how the rankings should be used, the report says the thing that connects it to everything above.

"The premium rate listed for a class will often differ from the rate that an individual employer would pay. Premium rates for an employer are adjusted based on the factors including the employer's experience rating, premium discounts, premium reductions"

The document that measures the price of workers compensation across the country states plainly that the number it publishes is not the number anybody pays, and names experience rating first among the reasons.

Where this is thin

Three accounts are three accounts. Every figure the contributors give is self-reported, appears in no public dataset, and describes businesses whose books nobody else has seen. Not one of the three gave an actual modification factor, so nothing here shows what any of their multipliers was, only what they say it did.

The Minnesota manual governs Minnesota. Attia advises across many states and Spisak's warehouse is not in Minnesota, so the manual is not the rule that priced either of their accounts. It is used here because the mechanism they both describe is written down somewhere public, and this is a place it is written down. Whether the equivalent rule in their states reads the same way was not checked this run, and no NCCI manual was read.

The Oregon study measures index rates, not premiums. Its own notes say the actual average premium rate for a jurisdiction will differ from the weighted index, that a different class selection or a different state's payroll weighting would produce different results, and that the figures are not adjusted for dividends paid back to employers. Setting Spisak's 180,000 dollars or Smith's 158 next to an index rate shows what each measure contains, not that one validates the other.

What survives all of that is narrow and reasonably firm. Two employers at opposite ends of the scale and one adviser who sees hundreds describe a price that moves for reasons other than the losses they had. The published rules for computing that price say the input is the reserve rather than the settlement, that fault is irrelevant, and that corrections run backwards only so far. The national study of what the cover costs says in its own notes that the rate it publishes is not the rate an employer pays, because experience rating changes it first.

Spisak spent on safety equipment, training and automation, and watched his claims fall and his premium rise. Attia moved an account by nearly half without touching the work at all. Both outcomes are what the arithmetic produces.