Romania’s ninth straight month of net business shrinkage came with its smallest headline loss since February. Look one layer down and the relief evaporates.

On the surface, August 2026 looks like the first piece of good news in a very long time. Romania registered 8,103 new businesses and lost 8,744 through suspensions, dissolutions and deregistrations. Net: minus 641 firms — the smallest net loss since February, in a nine-month run that began in December 2025.

That figure will be sold as moderation. It isn’t.

The pool shrank, so the loss shrank

The question that matters is whether August’s smaller net loss reflects genuinely fewer exits or simply a smaller base of registrations to lose against. The data answers unambiguously: both sides contracted, which is why the ratio looks better while the economy looks worse.

Start with the raw run-up in both sides of the ledger:

August registrations fell 26.7% from July’s 11,049, a month-over-month plunge of 2,946 firms. Exits fell in near lockstep — 12,079 to 8,744. The net gap narrowed not because fewer firms closed relative to the economy’s size, but because the whole pipeline slowed to a crawl.

The ratio that matters is still above 100

Here is the number nobody will lead with. Romania’s churn rate — exits as a share of registrations — sat at 107.91 in August. For every 100 businesses that opened in Romania last month, roughly 108 closed, dissolved, or were struck off.

Yes, that is down from 123.88 in April. But a churn rate above 100 is not a recovery metric. It is a contraction metric. The health ratio — registrations divided by exits — is 0.93. The ecosystem has now spent nine consecutive months below 1.0, and the cumulative damage since December 2025 stands at minus 16,756 businesses.

A company registrar that produced 16,756 fewer firms than it retired across nine months is not stabilizing. It is bleeding at a steadier drip.

“Worse than last year” undersells it

If you want the truly alarming comparison, don’t look at July. Look at August 2025.

Registrations for August 2026 are down 29.14% year-over-year — 8,103 against the 11,435 recorded a year earlier. The collapse is almost entirely a corporate-form story: SRL registrations cratered 41.95%, from 7,613 to 4,419. The 12-month moving average of registrations has now slipped to 12,627 from 12,904 a month earlier, a drop of 278 firms.

And the exits-side “improvement” everyone will cite? It is a year-over-year illusion worth puncturing. Dissolutions fell 31% versus August 2025 (4,635 to 3,198), deregistrations fell 19% (5,594 to 4,531), suspensions fell 31.42% (1,480 to 1,015). Superficially, fewer businesses died. But fewer businesses also existed to die — and the base they are being measured against shrank 29%. Subtract less from a smaller pile and the arithmetic looks kind. It isn’t.

The seasonality alibi doesn’t hold either

The obvious defense is that August is simply a quiet month. The data confirms August is the third-weakest calendar month on record, with mean registrations of 10,681 against March’s 13,289.

But that average was computed across ten years. August 2026 produced 8,103 — roughly 24% below the long-run August norm. This is not a seasonal lull. It is a seasonal lull sitting on top of a structural collapse. When the seasonal excuse applies to every one of nine consecutive failing months, it stops being an explanation and becomes a coping mechanism.

Where the damage is concentrated

The vulnerability map is stark. August’s registrations fell hardest in exactly the sectors that signal real economic activity:

• Hotels and restaurants: -52.26% year-over-year (597 to 285)
• Manufacturing: -46.71% (486 to 259)
• Other services: -38.98%
• Retail and wholesale trade: -38.11%
• Construction: -33.05%

The standouts holding up are thin. Transport and logistics fell only 6.02%, and information and communications dipped just 2.69% — but these are the sectors sitting at the top of the registration table, at 1,515 and 833 firms respectively. When the only sectors not collapsing are the ones absorbing the displaced, that is not diversification. That is a waiting room.

Geographically, it is a one-city story with a long tail of attrition. Bucharest generated 2,276 of August’s 8,103 registrations while also absorbing 980 of the country’s deregistrations. But the real signal is in the small counties: Tulcea down 48.44%, Harghita down 36.25%, Mehedinți down 33.72%, Satu Mare down 31.4%. Only Vaslui grew — by 5%, adding four firms. When your national growth map consists of Vaslui adding four businesses and everything else falling, the word “moderation” is doing an enormous amount of unearned work.

The verdict: continuity, not stabilization

Is this stability, or the absence of change dressed up as a finding? It is the latter.

Three things are true simultaneously, and only reporting one of them is misleading:

1. Exits are genuinely falling in absolute terms — dissolutions down 1,437 year-over-year. This is real.
2. Registrations are falling faster — down 3,332 year-over-year, with SRL formations in freefall.
3. Therefore the churn ratio improves while the net position remains negative for a ninth month and the cumulative toll passes 16,700 businesses.

A smaller loss is not a smaller problem when the loss is measured against a shrinking denominator. August 2026 did not moderate Romania’s business destruction. It simply ran the same destruction engine at lower RPM against thinner fuel — and produced a marginally less ugly number that will be mistaken for the turning of a corner.

Calling that a “moderation” would not just be optimistic. It would be arithmetically illiterate. The churn rate is still 107.91. The health ratio is still 0.93. The streak is still negative. Nothing has changed except the volume, and volume is not a verdict.