What Michigan employers will actually pay in 2027

Michigan’s insurance regulator requires every carrier to file its small-group rate change before the plan year begins, then publishes those filings with the number of covered lives behind each one. No license, no subscription, no vendor. It is one of the few genuinely public inputs into a renewal conversation.

Almost nobody reads it. What circulates instead is a single statewide average — and that average is built in a way that makes it misleading for any particular employer.

The headline: 9.6 per cent

Michigan’s approved small-group increase for 2026 was 11.1 per cent. The proposed figure for 2027 is 9.6 per cent. On its face, relief.

Look at what sits underneath. Eight carriers filed for the 2027 small-group market, and five of the eight went up, not down. The statewide figure falls anyway, because it is weighted by covered lives and one large carrier filed a low number against a very large block of members. The average moved. Most of the market did not.

The spread is the part worth carrying into a renewal meeting. Across those eight filings, 2027 small-group changes run from 7.0 per cent at the low end to 14.8 per cent at the high end — a gap of nearly eight points between one carrier’s filing and another’s, for the same market, in the same state, in the same year.

The split nobody labels

Michigan licenses health maintenance organizations and commercial insurers under separate statutory tracks, and each entity files its own rates. Every carrier family therefore appears in the table twice, under two corporate names — and the table never says which is which.

Separate them, weight each side by its own covered lives, and 2027 stops being one market:

Both halves reconcile exactly to the published total — the covered-lives counts add up with nobody dropped or double-counted.

Quote 9.6 per cent to an employer on an HMO and their renewal looks about a point and a half worse than their market actually was. Quote it to an employer on a PPO and it looks about a point and seven tenths better. Same sentence, opposite errors, turning on a product distinction the published table does not print.

And the Split Is Unstable

Run the identical calculation on the 2026 filings and the two halves come out at 11.10 and 11.05 per cent — no gap at all. The entire 2027 divergence comes from one carrier’s filing on a very large block of members. Anyone who learns this year’s shape and carries it forward will be confidently wrong.

Above fifty employees, none of this is your number

This is the limit that matters most, and it comes from the regulator in plain terms: large-group rates — over fifty — are not required to be filed at all, and the department states it has no authority over the rates charged under a self-funded plan.

So the filed table describes fully insured groups under fifty, and nobody else. Three limits travel with every figure in it: funding type, group size, and product type. Strip any one away and the number stops meaning what it appears to mean.

For a fully insured group over fifty, the closer reference is national employer trend, which for 2026 sat at 9 to 10 per cent before plan-design changes and roughly 6.5 to 7.6 per cent after them. That is a fourth consecutive elevated year, following a decade that ran near three.

The before-and-after distinction is the single most misquoted thing in this subject. It is a gap of two to three points, and the lower number already assumes the employer changed something — raised a deductible, moved a network, shifted contributions. Quoting the post-change figure as the market’s trend quietly credits an employer with savings they have not yet decided to make.

If you are self-funded, there is a third clock

The filed table does not apply to you at all, because you are not buying a filed rate. What moves instead is claims, administration and stop-loss, and they move independently.

Stop-loss premium rose 13.6 to 15.9 per cent for 2026 depending on deductible level — around five points faster than the previous cycle. The mechanism is leveraged trend: the attachment point stays where it is while claims above it keep growing, so the layer you are insuring inflates much faster than total spend. An employer whose claims rose ten per cent can open a stop-loss renewal in the thirties with nothing unusual having happened.

The question to take into your own renewal

Every number above is context. None of it is a verdict on your plan — and the reason reframes the conversation you are about to have.

A renewal blends two things: your group’s own claims experience, and the carrier’s book rate built from age, location and industry. The weight given to your own experience is called credibility, and it scales with the number of enrolled contracts — not employees. Full credibility is commonly set at a thousand enrolled contracts, with the book rate carrying the balance below that.

Work the arithmetic through and it changes what your increase can possibly mean:

• At a hundred enrolled contracts, roughly a tenth of your renewal reflects your own claims. Nine tenths is the book. A hard claims year barely shows — so anyone implying your increase is a verdict on your workforce is guessing, and the arithmetic says otherwise.
• At around five hundred, the market baseline and your own experience carry roughly equal weight. It is the only point on the scale where both halves genuinely matter at once.
• At a thousand and above, the renewal essentially is your own year, priced. A national trend figure quoted at you here is evidence about somebody else.

Two things follow. First, enrollment is a rate input, not just an HR statistic — two companies of identical headcount can sit at very different credibility depending purely on take-up, and thin participation is charged for separately on top of that. Second, and more useful immediately:

When we build a trend exhibit it always carries three lines, in this order: national, which sets the era; local, which sets the market and the product type; and your own number, which is the verdict. The client line means nothing without the two above it. The two above it are trivia without the client line. Fewer than three lines and the exhibit is not finished.