The sole-trader share of new Romanian business registrations fell to 39.8% in September 2026, down from a 2026 high of 41.8% in February. That looks like the beginning of a reversal. It is not. Measured against its own monthly volatility, September’s move is well inside the noise band, and the composition of Romania’s business register held essentially where it has been all year — an economy still tilted about four-tenths toward unincorporated forms, with the incorporated SRL holding near 56%.

The arithmetic of the “retreat”

September produced 6,912 registrations in total. Of those, 2,751 were PFA sole-trader registrations and 3,878 were SRL incorporations. That is a PFA share of 39.80% and an SRL share of 56.11%.

Across the eleven months that can be reconstructed for 2026, the highest PFA share was February’s 41.82% — 5,495 of 13,139 registrations — and July came close at 42.19%, with 4,662 of 11,049. September’s 39.80% is roughly two percentage points down from the February high.

But the monthly PFA share series for 2026 is: 40.2% in January, 41.8% in February, 40.2% in March, 39.0% in April, 38.8% in May, 40.2% in June, 42.2% in July, 41.3% in August, 39.8% in September. That is a band of 38.8% to 42.2% — a range of 3.4 percentage points, swinging both directions twice within nine months. September’s fall from July’s 42.2% is a 2.4-point move, and August had already given back 0.9 points of it.

Put plainly: the month-to-month variation in this series is larger than the distance between the recent high and the current reading. A single month’s move of this size has already happened three times this year without marking a turning point. Calling September a reversal requires evidence September does not provide.

What September does tell us

There is a real compositional fact here, but it is not a September event — it is a level fact that September confirmed.

A year earlier, September 2025 registered 14,815 businesses, of which 9,343 were SRLs and 4,938 were PFAs: an SRL share of 63.07% and a PFA share of 33.33%. September 2026’s SRL share of 56.11% is 7.0 percentage points lower, and its PFA share 6.5 points higher, than the same calendar month a year ago. That comparison is not distorted by the seasonal position of the month, because both readings come from September.

This is the more honest framing of what has changed. The incorporated form has ceased to be the automatic default for a new Romanian business: in September 2026, more than two in five new registrations were not SRLs, where a year earlier fewer than one in three were.

The drift is also visible over a longer horizon, though with a caution. September 2023 showed an SRL share of 64.7% and a PFA share of 27.9%; September 2022 showed 65.2% and 21.8%, though those older years carried much heavier IF and CA counts — 1,147 and 350 respectively in September 2022 — that have since largely disappeared from the register. The IF line fell from 262 in December 2023 to 15 in September 2026, and CA from 115 to 15. Part of the apparent shift in the two-party SRL/PFA split is a reallocation away from those older intermediate forms rather than a pure defection from incorporation. Comparisons reaching back before roughly 2024 should be read with that in mind.

The volatility is concentrated in the small months

One reason not to over-read September’s ratio: the absolute numbers shrank a lot, and when the denominator shrinks, the ratio gets jumpy. September’s 6,912 registrations compare with 11,049 in July — a fall of 37.4% — and August came in at 8,103.

That contraction is not a compositional story, it is a volume story. What matters here is the mechanical consequence: a 3.4-point share band on a base of 6,000–7,000 registrations is a fluctuation of a couple of hundred firms. On a base of 13,000 it would not register at all.

Both the incorporated and the unincorporated form declined year on year in September, and the SRL declined faster in percentage terms — down 58.5% from 9,343 to 3,878, against a 44.3% fall for the PFA from 4,938 to 2,751. That is the mechanism behind the share move: not more sole traders, but fewer incorporations. The absolute number of new sole traders fell by 2,187; the PFA share still rose because the SRL count fell by 5,465.

This distinction matters for interpreting the year-on-year figure. The rise in the PFA share is arithmetically a shrinkage effect, not a growth effect. Nothing in this month’s data shows a wave of people choosing the sole-trader form; it shows fewer people choosing the incorporated one, against a smaller overall pool.

Exits are running ahead of entries, and the exit side is shrinking too

Composition can only be read alongside churn. September recorded 7,629 business exits against 6,912 registrations — 845 suspensions, 2,875 dissolutions and 3,909 deregistrations — for a churn ratio of 110.4% and net growth of −717. September was the tenth month in an unbroken run of negative net growth that began in December 2025, with a cumulative total across the run of −17,473. That measurement window begins only in July 2024, the earliest month with reliable exit data, so the run is not comparable to the full historical record. Within the ten months, September’s −717 ranks third mildest.

The relevant point for a composition argument is that exits are falling nearly as fast as entries. September’s suspensions were 46.8% below September 2025’s, its dissolutions 46.2% below and its deregistrations 35.1% below. An economy shedding incorporations and shedding closures at similar rates is a quiet one, not a restructuring one. A genuine composition change would show differential exit rates by legal form; that breakdown is not available, and this report does not claim it.

Where the structure actually differs: geography

If you want to find real compositional variation, it is regional rather than temporal. Bucharest alone accounted for 1,779 of September’s 6,912 registrations, 25.7% of the national total, followed by Ilfov at 510 (7.4%), Timiș at 337 (4.9%) and Cluj at 330 (4.8%).

Cluj’s September cohort shows 191 SRLs and 136 PFAs — a PFA share of 41.2%, above the national 39.8% but built on 330 registrations. Where county-level data is available, the deviations from the national split are modest, and with month bases in the low hundreds a single month’s county ratio carries little weight.

The month’s industry distribution is led by transport and storage at 1,181, wholesale and retail at 1,004, information and communications at 737 and professional, scientific and technical activities at 722. Notably, information and communications and professional services — the two sectors where the incorporated form has the strongest conceptual case, given liability and contracting requirements — fell least year on year, at −29.9% and −50.5%, against transport at −45.1% and wholesale and retail at −52.6%. The deepest contractions were in hotels and restaurants (−74.9%), cultural and recreational activities (−74.6%) and manufacturing (−68.7%). A composition reading of that pattern would be speculative on one month’s data, but the relative resilience of the white-collar sectors is at least consistent with the incorporated core of the register eroding more slowly than its periphery.

The verdict

September’s entity mix is monthly noise around a level that has been remarkably stable through 2026. The PFA share’s move from 42.2% in July to 39.8% is smaller than the peak-to-trough range the same series covered between March and May, and no larger than its swings in other months on much larger bases.

What is not noise is the year-on-year comparison: an SRL share down seven points on September 2025, driven by a 58.5% fall in incorporations against a 44.3% fall in sole-trader registrations. That is a real change in the composition of new Romanian business formation, but its cause is a weaker top line rather than a stronger preference for any particular legal form — the same SRL-specific rather than broad pattern the market has been showing since mid-2026, now visible in the register’s structure.

September neither continues the drift nor reverses it. It sits inside the band, and the drift itself needs a longer view to be measured than a monthly ratio on a shrinking base can supply.