September has never been a strong month for Romanian company formation — it ranks ninth of twelve by ten-year average, at 11,385 registrations. That alone would make a soft September unremarkable. But the gap between what September normally produces and what September 2026 produced is far too wide to be absorbed by seasonal variation: 6,912 registrations against a September norm of 11,385. That is a shortfall of 39% against the month’s own long-run mean. September 2026 cannot be read as an ordinary September.

The same-month record, laid out

Using only Septembers, the series reads as follows. In 2017–2019 September sat in a stable band of 9,585–9,705 registrations. It then rose sharply — 12,138 (2020), 13,140 (2021), a record 15,951 (2022) — before settling back into the 9,500–14,800 range across 2023–2025: 12,343; 9,585; 14,815. All ten observations in this window fall at or above 9,585.

September 2026, at 6,912, ranks last of the ten Septembers from 2017 to 2026 and sits 28% below the lowest of the other nine (September 2018 and September 2024, both 9,585). It is the only September in the ten-year window below 9,000.

The 2013–2015 figures visible in the register run to six digits — 100,146; 100,727; 79,084 — and are plainly a different recording basis, not a comparable volume of nine separate monthly readings. They are excluded here for that reason, and their exclusion makes the 2017–2026 comparison the meaningful one.

Why “seasonal” does not explain it

The seasonal test is straightforward: if September’s weakness were the normal rhythm of the calendar, the month-to-month move out of August should resemble previous years. It does not. August 2026 recorded 8,103 registrations, and September’s 6,912 represented a fall of 14.7%. Since 2017, September has risen over August in every single year — from 9,674 to 12,138 in the pandemic year 2020, and from 9,585 to 14,815 in 2025. A September that contracts against August is unprecedented in the ten-year window.

The year-over-year picture points the same way and is far more severe than anything else in 2026. September 2025 recorded 14,815 registrations; September 2026 recorded 6,912, a change of −53.3%. That is not a continuation of a gentle drift: the same comparison was positive in January (+27.7%), February (+8.6%), March (+14.5%) and April (+10.1%), and only mildly negative in May (−3.5%) and June (−4.2%). The break comes in the second half: −27.2% in July, −29.1% in August, and −53.3% in September. Whatever is happening, it steepened sharply and abruptly at the end of the third quarter.

The 12-month moving average confirms the same point without relying on a single month. It stood at 13,248 in June and 11,968 in September, down 659 from the prior reading. September’s own value of 6,912 is 42% below that rolling average — a dispersion too wide to be ordinary monthly noise.

The one honest caveat is data completeness. September’s total is less than half of August’s, and October 2026 returns no records at all. Reporting lags can depress the most recent month in any register, so the size of the September drop should be treated as provisional until the month is fully settled. But the year-over-year gap amounts to 7,903 fewer registrations, and the July and August declines were already recorded before September arrived.

The entity mix is consistent with September’s usual composition; the scale is not

The entity mix in September 2026 is broadly what September normally looks like — which is itself notable, because the aggregate is not.

The incorporated SRL form accounted for 3,878 of September’s registrations, or 56.1% of the total. Sole-trader PFA registrations came to 2,751, or 39.8%. Compare that to the same month a year earlier, when SRLs were 9,343 of 14,815 (63.1%) and PFAs 4,938 of 14,815 (33.3%). The shift is real but moderate: roughly seven percentage points moved from the incorporated form toward the sole-trader form, with the remainder in small authorized-person and family-enterprise categories.

That is a far smaller compositional shift than the headline fall would suggest. Every large entity category contracted sharply and roughly in parallel year-over-year — SRL registrations down 58.5%, PFA down 44.3%, individual-enterprise II down 52.4% — while only the tiny categories moved the other way on trivial bases: CA at 15 registrations and SNC at 9. A month in which the mix holds roughly in place while every major component falls proportionally is a month of broad contraction in formation volume, not a rearrangement of the business fabric.

September 2026 also carries the third-mildest net position of the current ten-month run of negative business formation, at −717. That streak began in December 2025 and has accumulated −17,473 over ten months, measured from July 2024, the earliest month with reliable exit data. Lifecycle events in September totalled 7,629 exits against 6,912 registrations, a churn rate of 110.4%. Exit activity itself is down year-over-year — deregistrations 3,909 against 6,025, dissolutions 2,875 against 5,339 — so the negative net is being driven by the collapse in entries rather than by a surge of closures.

Sectorally, the largest industries kept their relative order: transport and warehousing led with 1,181 registrations, followed by wholesale and retail trade at 1,004, information and communications at 737, and professional, scientific and technical activities at 722. Every one of these fell by 30% to 65% year-over-year, with manufacturing and “other services” down the most.

Geographically, Bucharest accounted for 1,779 registrations, or 25.7% of the national total, with Ilfov, Timiș, Cluj and Iași following. The decline is not confined to the capital’s satellites: the sharpest percentage drops appear in smaller counties such as Covasna and Harghita, but the volume-weighted fall is national.

What the record supports

Two conclusions survive the data. First, September 2026 is not an in-range September. It is the lowest September of the ten-year comparison by a margin of nearly 2,700 registrations over the next-lowest, it is the first September below 9,000, and it is the first September since 2017 to fall rather than rise from August. Placement against the multi-year record, not against last month or last year, is what makes that unambiguous.

Second, the entity mix is consistent with September’s customary composition, with a modest tilt toward sole-trader forms. The month’s problem is scale, not structure. That distinction matters for how the 2026 record is read: the sequential deterioration from a positive first quarter to a 53% year-over-year decline in September describes an accelerating contraction in the number of businesses being registered, arriving on top of — not instead of — the ten-month run of negative net formation. Whether September is the bottom of that deterioration or an outlier within it cannot be determined from this data, and no claim either way is warranted here.