The Legal Entity Identifier system asks every registered entity to name the parent that consolidates its accounts. Of the 748 entities with a Philippine legal address, 192 answer. 515 file a coded reason instead — and 50.5% of those reasons are not a refusal at all. They are a declaration that no such parent exists.
Naming a parent is not compulsory. Filing a coded reason for not naming one is. So every entity sits in exactly one of three states per parent kind, and the interesting number is not how many owners are known — it is how the other two states divide.
168 entities name the company directly above them. 538 file a reason instead, and 42 record neither.
192 name the company at the top of the chain — more than name a direct parent, because an entity with no intermediate holding company still has an owner above it.
entities record neither a parent nor a reason at the ultimate level, about one in eighteen. The obligation is working; it is the content of the answers that leaves the graph open.
5 codes account for every declined parent, and the order they fall in is the finding. The graph is open mostly because these entities report no parent at all; after that because the owner is the wrong kind of thing to hold an identifier. Outright non-disclosure is the smallest group of the three.
260 entities say nobody consolidates their accounts, or that no controlling person is known. For these the graph is not incomplete — it is complete, and the entity is a root.
203 say the parent is a natural person, or a company that holds no LEI. The owner exists and is known; there is simply no identifier to point at, so the edge cannot be drawn.
52 cite non-disclosure. The category a reader assumes dominates this data accounts for one declined parent in ten.
Those three groupings are this page's reading, not GLEIF's. The codes are the
registry's own closed vocabulary and are reported verbatim; what each code
implies ships as a column in
data/ph-lei/ph_lei_reason.csv so it can be argued with, and a check
fails the build the day a code arrives that this map has never seen. That is the
difference between an interpretation you can audit and one buried in a chart
colour.
The 192 declared ultimate parents sit in 30 jurisdictions, and 84.4% of them are abroad. That is roughly what an LEI is for: it is usually obtained because a foreign counterparty asked for one.
of the named ultimate parents are themselves Philippine-registered — third place, just behind the United States on 41 and Germany. Those edges close inside the country instead of pointing abroad.
Philippine entities roll up to Aboitiz Equity Ventures Inc., and no other parent in this slice holds more. The biggest owner of Philippine LEI holders is Philippine.
parents hold more than one Philippine entity. The rest of the graph is a wide, flat fan rather than a few deep groups.
GLEIF publishes a per-record modification history — field, old value, new value, date. It is slowly-changing-dimension capture served directly, which is why nothing here diffs the 516 MB daily snapshot. It also contains a trap.
Every field of every record appears once with modification type
INITIAL: the moment the registry first wrote it down. That is
32,582 of the feed's 78,818 rows, or
41.3%. A change-capture job that does not
subtract them fires on the whole archive the first time it runs and calls the
result churn — 78,818 changes where there are in fact
46,236.
The feed also says which record changed: 52,623 rows describe the entity itself, 22,436 its relationship declarations, and 3,759 its reporting exceptions. The refusal to name an owner has a revision history of its own.
Strip the first loads out and rank what is left by field. The top of that list is not ownership, or names, or addresses.
changes to LastUpdateDate on the
entity record alone. The most-changed field in the feed is the field that
records that the record changed, so an alert built on "this record moved" is
mostly an alert about itself.
recorded changes is the median across all 748 entities, over a feed running from 2018 to 2026.
entities have no recorded change at all beyond their first load — registered, and then still.
Four limits, and the first one is the one most likely to be misread.
The LEI parent is defined by accounting consolidation — who includes this entity in their financial statements. That is a narrower question than who ultimately benefits, and an entity that honestly reports no consolidating parent may still have a very definite owner.
These are the Philippine entities that obtained an LEI, usually because a counterparty or regulator required one. They are a slice of the country's registered companies, not a sample of them, and nothing here generalises to the rest.
44.8% of these records carry no region in their legal address at all. The country is always there; the province often is not, so any map below national level would be drawn on barely half the data.
The modification history starts at 2018, six years after the LEI system did. Anything older is outside it, so "first recorded" is not the same as "first true".
LastUpdateDate with
6,496 changes — the record noting that it
changed.
Every figure on this page traces to one of these, through a CSV in
data/ph-lei/. Each is checked against its source query on every
build.
The 748 legal entities whose registered legal address is in the Philippines, with their legal name, city, region, legal form, entity status and registration status — Keyless: a plain GET, no account, no token, no form. Filtered server-side on filter[entity.legalAddress.country]=PH and paged at the API's maximum of 200 records. The paged count is asserted against the total the API reports, so a short page fails the build instead of quietly shrinking the country
The licence this page relies on, quoted in full: “The data available through the Access Service are provided under the CC0 licence” — CC0 is a public-domain dedication: no attribution condition, no non-commercial condition, no registration and no fee. The licence artifact itself was fetched rather than a documentation page describing it, because this site's rule is public domain or CC0 only and a docs page is not the instrument
Which entities name a direct or ultimate accounting-consolidation parent, and the parent's own LEI, legal name and country — A 404 here is data, not an error: it means no parent is recorded. The pipeline caches the 404 as an answer so a re-run does not ask again, and only then asks the exception endpoint for the reason. No bulk filter exists for these relationships — every variant tried returns 404 — so they are read one entity at a time
The coded reason an entity gives for not naming a parent, and the category of parent the exception applies to — Naming a parent is not compulsory; filing a coded reason instead is. These codes are GLEIF's own closed vocabulary and are reported verbatim. The grouping of those codes into what each one asserts is this page's reading, published as a column in ph_lei_reason.csv and checked, so a code nobody has classified fails the build
Every recorded change to an entity's LEI record, its relationship records and its reporting exceptions — the field, its old value, its new value and the date — A slowly-changing-dimension feed served directly, which is why this page does not diff daily files. Its first-load rows are typed INITIAL and are not changes; every change figure on the page is computed after removing them
Would have provided the same records as a daily full snapshot, roughly 516 MB per day — Not on licence — the golden copy is CC0 too — but on footprint and on method. Reconstructing a change history from it means storing and diffing consecutive full snapshots; the field-modifications endpoint already publishes the differences, so the 516 MB buys nothing this page needs
Would have returned every Philippine-rooted parent relationship in one filtered query instead of two requests per entity — No such filter exists: every documented and guessed variant answers 404, and page[size] caps at 200. The traversal below is what the absence of a bulk filter costs, and it is published as a figure on the page rather than hidden in the method
I wrote a companion post — “Who Owns This Company? Mostly, Nobody You Can Point At.” — in simple, everyday words.