September 2026 produced 6,912 registrations against 7,629 lifecycle exits — a net loss of 717 businesses and an exit-to-registration ratio of 110.37 exits per 100 registrations. The streak of negative months now stands at ten, running back to December 2025. What that ten-month figure does and does not establish is the subject of this piece, and the honest answer is narrower than the headline number suggests.

The streak is real, but its length is a within-window artefact

The run of net-negative months now stands at ten, beginning in December 2025, with a cumulative net loss of 17,473 businesses. December 2025 was the extreme month at −4,292; February 2026 was the mildest at −119.

The same series carries a limitation that must be reported rather than buried: it is measured only from July 2024, the earliest month with trustworthy exit data, and is not comparable to the full historical record — it cannot be called an all-time high or the longest ever. Registration data stretches back to 2013, but the exit side of the ledger — suspensions, dissolutions, deregistrations — only becomes dependable from mid-2024. A net-growth figure requires both sides. So “ten months” is not a claim about Romanian business history; it is a statement about how far back the computable series reaches. Whether an equivalent run occurred in 2016, 2019 or 2022 is simply unknowable with this data. Reporting the streak as any kind of record would be a category error.

The truncation point matters differently than it might appear. The ten-month run does not reach the edge of the trustworthy window — it ends comfortably inside a window that is itself only about 27 months long. The constraint is not that the streak is cut short by missing data; it is that even a streak terminating well within the observable period cannot be benchmarked against anything earlier.

September’s ratio sits near the mild end of the 27-month distribution

Testing whether September’s exit-to-registration ratio is inside or outside the distribution of the trustworthy window required reconstructing that distribution month by month. The result cuts against a dramatic reading.

Within the window from July 2024 to September 2026, monthly exit-to-registration ratios (total exits ÷ registrations × 100) ranged from 80.84 in October 2025 to 197.29 in December 2024. September 2026’s ratio of 110.37 sits close to the middle of that range and, notably, below several months that preceded the current streak. December 2024 (197.29), January 2025 (133.68) and August 2024 (133.83) were all considerably worse than September 2026 — and one of them, January 2025, was a net-negative month that did not begin a streak at all.

The month-over-month comparison is close to a flat line. August 2026 carried a ratio of 107.91 on 8,744 exits against 8,103 registrations; September’s ratio of 110.37 is marginally higher because registrations fell 14.7% while exits fell faster in absolute terms, from 8,744 to 7,629. Net growth moved from −641 to −717, and the health ratio eased from 0.93 to 0.91.

What the September figures actually describe

The composition of September’s exits is worth stating precisely, because the three event types behave differently. Deregistrations were the largest category at 3,909, dissolutions 2,875 and suspensions 845. Deregistrations and dissolutions represent permanent exits; suspensions are temporary and arguably should not be aggregated with them at all — a definitional choice embedded in the total exits field that inflates the ratio by roughly 11 percentage points this month. Readers comparing separation rates across sources should check which categories are combined.

The year-over-year picture is where September is genuinely unusual, and not in the direction the streak implies. Every exit category fell sharply against September 2025: suspensions down 46.75%, dissolutions down 46.15%, deregistrations down 35.12%. Registrations, however, fell faster still — 53.34%, from 14,815 to 6,912 — with SRL formations down 58.49% to 3,878 and PFA down 44.29% to 2,751. September 2026 is therefore not a month of unusually heavy closures. It is a month of unusually thin formation, against a closure base that has contracted even more.

Some of that thinning is seasonal. September is a weak registration month in the structural pattern, ranking ninth of twelve with a mean of 11,385. But no observed September has previously fallen near 6,912, which the twelve-month moving average of 11,968 makes plain; September 2026 came in at 58% of that trend. Seasonal weakness explains part of the level; it does not explain all of it.

Where the stress appears, and where it doesn’t

Entity-level data shows the burden falling unevenly across legal forms. SRL formations of 3,878 contracted faster year-over-year (−58.49%) than PFA formations of 2,751 (−44.29%), meaning the incorporated form typically associated with larger ventures is thinning faster than sole-trader formation. Capital-marked forms are numerically negligible: six SA registrations, fifteen CA, three SC.

Geographically, exits concentrate where the registrations do — but the relationship is not uniform. Bucharest recorded 1,379 exits against 1,779 registrations for a ratio of 77.52, the lowest among the largest counties in this month’s data; Ilfov was lower still at 71.96, while Constanța reached 132.94 and Cluj 119.39. At the far end, Vaslui recorded 99 exits against 34 registrations, a ratio of 291.18, and Tulcea 244.12 — small absolute numbers in both cases, which makes their ratios volatile rather than necessarily alarming.

Sectoral registration declines were broad rather than concentrated — manufacturing −68.74%, construction −53.55%, trade −52.62%, transport −45.10% — though transport and warehousing still produced the largest absolute number of new registrations at 1,181. Caution is warranted here: the data supplies registrations by sector, not exits by sector, so sectoral vulnerability cannot be identified from this dataset. Any claim about which industries are shedding businesses most is unsupported by what these tools return.

The conclusion

Three findings survive scrutiny. First, Romania has recorded ten consecutive months in which lifecycle exits outnumbered registrations, cumulatively subtracting 17,473 from the business stock — a real and consistent pattern within the observed period. Second, that period is roughly 27 months long, beginning July 2024, so the ten-month run cannot be characterised as a record of any kind; it is a within-window finding and should always be reported as one. Third, September’s exit-to-registration ratio of 110.37 is unremarkable for the window — inside the distribution, closer to the mild end than the severe end, and worse months occurred before the streak began.

What the data cannot establish is equally clear: whether the current configuration of weak formation and modest exits is historically unusual, because there is no trustworthy pre-2024 exit series to compare it against; and which sectors or counties are driving exits, because exits are not published at that granularity here. The one thing that can be said with confidence about the health of the ecosystem is directional and modest: formation is running well below trend, exits are running below last year’s level, and the balance remains negative for a tenth month. That is stress on the entry side far more than the exit side — a distinction the streak count alone conceals.