Swiss commercial registry publications, classified: monthly panel
Monthly counts of Swiss commercial registry publications from 2018-09 to 2026-08, by canton and by the kind of change each one announces, with capital movements in CHF. 18,668 rows, free under CC BY 4.0.
Download
| File | Size | SHA-256 |
|---|---|---|
| shab_registry_panel.csv | 745 kB | 42b039a32d0d53eb… |
| shab_registry_panel.parquet | 51 kB | 571d335421a38295… |
| data_dictionary.md | 1 kB | ddb73c2e07a5edaa… |
| METHODOLOGY.md | 5 kB | a8048853e847069b… |
| datapackage.json | 3 kB | ad3254cecd0837f1… |
| LICENSE.txt | 640 B | 5ce64fd18e93415a… |
Citing this dataset
Prospex (2026). Swiss commercial registry publications, classified: monthly panel, release 2026.09. Derived from the Swiss Official Gazette of Commerce (SHAB/FOSC). CC BY 4.0. https://prospex.ch/open-data/swiss-registry-panel/
Attribution is a condition of the licence. A link to this page is the attribution we ask for.
What one row is
One row is one month, one canton and one kind of registry change. The file is in long format and it is sparse: a combination that produced no publications has no row at all, which means the same thing as a count of zero.
| Column | Type | Description |
|---|---|---|
month | date | First day of the calendar month in which the publication appeared in the SHAB. |
canton | string | Two-letter code of the canton whose register office published the entry. XX where the publication carries no canton. |
event_type | string | Prospex's classification of what the publication announces. |
event_count | integer | Publications classified into this month, canton and type. |
n_events_with_amount | integer | Of those, how many stated both a prior and a new nominal capital. Zero outside CAPITAL_INCREASED. |
capital_delta_sum_chf | number | Sum of new minus prior nominal capital, in CHF, over the events counted in n_events_with_amount. Empty when withheld. |
capital_delta_median_chf | number | Median of the same differences, in CHF. Even-length samples take the mean of the two middle values. Empty when withheld. |
capital_suppressed | boolean | True where fewer than three events carried an amount and the two capital measures are withheld. |
Where it comes from
Every count is derived from the Swiss Official Gazette of Commerce (SHAB/FOSC), the official publication of record for the Swiss commercial register. Prospex ingests the gazette and classifies each publication into one of the event types below.
This page reports a classification of official publications. The Swiss authorities publish the gazette; the event type attached to a publication is Prospex's reading of it.
INCORPORATIONCAPITAL_INCREASEDSEAT_MOVEDADDRESS_CHANGEDMERGERBRANCH_CREATEDDELETEDLIQUIDATION: the rate moves from about 85 per 1,000 in 2019 to about 48 in 2021 and back to about 75 in 2026, and no parser-version boundary explains it.
Event types withheld
The event types below are withheld, because their measured rate tracks changes in our parser instead of changes in the register.
OFFICERS_CHANGED: Rate is flat at 62–78 per 1,000 publications through 2025-01, then 120.8, then 361–381. The jump is our own reprocessing of officer entries, not a change in Swiss corporate behaviour.NAME_CHANGED: Rate is exactly 0.0 before 2025-07 because no earlier parser emitted the type. Publishing it would show a trend that is entirely an artefact of our own code.
Limitations
Capital movements are a lower bound. A capital figure is recorded only where the publication states both a prior and a new nominal capital and the new figure is the larger. Only increases are emitted. Any average must take n_events_with_amount as its denominator.
Small cells withhold their capital figures. Where fewer than three events in a cell carried an amount, the two capital columns are empty and capital_suppressed is true. A mean over one event describes a single company. event_count and n_events_with_amount are always populated.
The month is the month of publication. The gazette publishes a few days after the register entry takes effect, so a change made at the end of a month can appear in the next one.
Legal form is unavailable for struck-off companies. Legal form is available only from a company's current register profile, and a company that has been struck off no longer has one, so the missing values correlate with the very thing being measured. This release carries no legal-form dimension.
How the period was chosen
The period follows from four checks. A calendar month enters the dataset only if it passes all of them:
- The ingestion cursor has passed the end of the month.
- No failed ingestion run covering it was left unsuperseded.
- Every working day in it carries at least one publication, except the confirmed federal holidays.
- Fewer than 0.5% of its publications failed to parse.
The published period is the longest unbroken run of passing months. This release covers 96 months and holds 18,668 rows. The full method, including every month considered and rejected, is in METHODOLOGY.md above.
Mirrors
The same files are on Hugging Face. This page is the canonical version.