Property talk:P1320

I've just been looking into how this is constructed. At the moment, it links to a specific company - ie, a single legal entity. For AstraZeneca, for example, it uses gb/02723534. However, this is one of nine companies which are part of a broader group, and it's really the broader group we think of when we say "AstraZeneca" - the UK main company, the Spanish, Dutch and Slovenian subsdiaries, the UK financial holding company, etc, are all part of the same thing.

Would it be better to have this identifier use the OpenCorporates "corporate grouping" concept? They describe this as a "way of grouping of companies without having to say exactly what legal form that relationship takes (e.g. it may be a subsidiary of a subsidiary). In short, it's what most humans (i.e. non tax-lawyers) think of when they think of a large company.". This matches very well to the way Wikipedia/Wikidata tends to handle companies - the general entity rather than the individual legal elements.

AstraZeneca would then link to AstraZeneca - which links to the WP AstraZeneca page, as it happens. Thoughts? AstraZeneca is a relatively simple case, but someone like Stagecoach links 86 seperate OC entries, and large international conglomerates could involve hundreds. Andrew Gray (talk) 11:47, 17 January 2015 (UTC)Reply

@Gangle, Addshore, LydiaPintscher, John Vandenberg: Andrew Gray (talk) 11:48, 17 January 2015 (UTC)Reply

I am guessing the corporate_groupings are stable identifiers? ·addshore· ^{talk to me!} 11:52, 17 January 2015 (UTC)Reply

Once created, I think so. However, they do need to be manually created (unlike the individual company registrations, which are always there) Andrew Gray (talk) 12:19, 17 January 2015 (UTC)Reply

Perhaps we should consider using both? Sometimes our entries will relate to an OC grouping (which, BTW, should have a 1-1 relation with Wikipedia articles; note the Wikipedia link on OC's AstraZeneca grouping page), sometimes individual components (and sometimes we're describing an intermediate type, with no direct OC equivalent). Not every company on OC or in Wikidata will have an equivalent OC group or Wikipedia article. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:18, 18 January 2015 (UTC)Reply

Individual company entries have a specific meaning in OpenCorporates, in that they are an assertion (with proof) that a company exists in a corporate register that has legal force. Corporate groupings, by contrast, are community curated and only lightly moderated. I agree that Wikipedia/Wikidata entries tend to refer to a "folk" entity (which has no legal force), and this entity corresponds with the OpenCorporates grouping concept. However, I'm not sure what the right solution is, because while a grouping may have a stable identifier, its members may not. How about locating the ultimate controlling entity and linking to that? (I work at OpenCorporates BTW) Gangle (talk) 07:59, 19 January 2015 (UTC)Reply

I think the "ultimate entity" is a sensible one, and would also help for the cases that Andy notes where we have articles on different parts of a group - eg Unilever (Q157062) but also Hindustan Unilever (Q1619376), Unilever Australasia (Q7885206), etc; we can link to the top of "that bit". However, is it easy to identify this in OpenCorporates at the moment? Andrew Gray (talk) 21:29, 19 January 2015 (UTC)Reply

Very late reply: the ultimate entity is easy to identify when we have it -- but our control data is quite sparse at the moment. Here's an example for a Microsoft subsidiary -- see "controlling company" near the top. Gangle (talk)

So to summarise the situation from an OpenCorporates perspective: on the one hand, there is a good fit with Corporate Groupings, because these are folksonomies curated by our community and as such tend to pertain to "Microsoft-the-abstract-thing-people-mean-when-they-talk-about-Microsoft". And this folk meaning is what people tend to have in mind when using Wikipedia to look for information about a company. On they other hand, our Corporate Groupings are not curated with a high degree of QA; whereas our core data of legal entities is. Also, when someone uses the folk sense of Microsoft, if you tried to pin them down, they would probably agree that what they *meant* was the ultimate controlling company: the one to which money and power ultimately flows. I would tend to say the ultimate controlling company is the correct entity to link to for these reasons, but as I mentioned above, our control data is relatively sparse right now. It's getting better all the time, but there are still lots of gaps. We've recently added weighting to our search results to rank controlling companies higher, but it's experimental so our default search is still sorted alphabetically (see "advanced options" in search results to get sorting by score). I'd suggest the way forward is to try going with ultimate controlling companies, and do an analysis of the gaps for notable Wikidata entries for which the ultimate controlling company is not clear; then we could look at ways to prioritise sourcing this data for OpenCorporates. Gangle (talk)

This sounds reasonable - "ultimate controlling companies" would also let us highlight the subsidiary parts which have identified entities, eg subsidiaries or acquired companies, by linking to the controlling company of that particular bit. Is there an obvious way to identify "ultimate controlling companies" when looking at the OC data? It seems reasonably human-obvious that, eg, this is Microsoft - but not sure how a script would be able to figure that out. Andrew Gray (talk) 12:58, 27 October 2015 (UTC)Reply

It just occurred to me that the second part of these ids (after the /) is identical to the corresponding id in the national company register. For instance OpenCorporates ID (P1320)fr/527678262 corresponds to SIREN number (P1616) "527678262" , and OpenCorporates ID (P1320)gb/02906991 corresponds to Companies House company ID (P2622)02906991.

• Is there anybody actively deriving national ids from OC ids and conversely?
• Do we have a mapping somewhere of which properties correspond to which OC prefixes?
• If not, how could we maintain that mapping in Wikidata itself? Maybe we could record the OC prefix on the properties themselves with catalog code (P528) or something similar?

− Pintoch (talk) 08:44, 26 June 2018 (UTC)Reply

I have done the derivation mentioned above for a bunch of jurisdictions and made a table to map OpenCorporates prefixes to Wikidata properties. That suggests a lot of new properties to create for each jurisdiction where the registry is open enough. OpenCorporates has an OpenRefine reconciliation interface that can be used to discover ids of existing company items in the national registries. − Pintoch (talk) 14:00, 29 June 2018 (UTC)Reply

@Pintoch: I've only just read about this property, and have converted Australia (Q408) in both directions, as well as setting up a complex constraint on Australian Company Number (P3549) to keep them aligned. I'll look into getting a set from the OpenCorporate API for matching the ID. I assume this is an ok use of the API @Gangle:, even though Wikidata is CC0? OC is generally in favour of Wikidata integration? --99of9 (talk) 10:09, 14 December 2021 (UTC)Reply

@99of9: Great! Yes importing ids is fine. − Pintoch (talk) 13:42, 17 December 2021 (UTC)Reply

I am aware of @Mormegil: occasionally adding OpenCorporates ID (P1320) based on Czech Registration ID (IČO) (P4156). It would be nice to have a bot doing that regularly, in both direction and for all relevant national properties.

I am not aware of any complete mapping, but feel free to use Wikidata:WikiProject Companies subpage for this purpose. All existing national properties are listed here Wikidata:WikiProject_Companies/Properties#Country-specific. Wikidata:WikiProject Companies/Country-specific has opencorporates column as well, but its country coverage is limited at the moment.--Jklamo (talk) 15:35, 17 December 2021 (UTC)Reply