openownership
Feature: Interest modelling
[This ticket helps track progress towards developing a particular feature in BODS where changes or revisions to the standard may be required. It should be placed on the BODS Feature Tracker, under the relevant status column.
See Feature development in BODS in the Handbook.
The title of this GitHub ticket should be 'Feature: XXXXX' where XXXXX is the feature name below. The information in this first post on the thread should be updated as necessary so that it holds up-to-date information. Comments on this ticket can be used to help track high-level work towards this feature or to refine this set of information.]
Feature name: Interest modelling
Feature background
Briefly describe the purpose of this feature
In BODS, the different ways that ownership or control can be exercised are called 'interests'. They exist between a subject (a company or other corporate vehicle) and an interested party (an individual, or a corporate vehicle) and interest types are captured in a codelist. The challenge for BODS is to represent the myriad mechanisms and channels of influence that individuals and entities use to benefit from and to control other entities. There are two dimensions to this challenge. Firstly, there is a challenge with coverage: how can BODS represent the full extent and scope of interests globally? Second, there is a challenge with modelling: what level of detail should the standard aim for, and to what degree should BODS try to standardise representation of diverse interests from different jurisdictions?
In BODS v0.3, the interest type codelist was updated to represent a broader range of known and unknown interest types. This relates to the first dimension of the challenge related to coverage. However, a number of modelling issues remain, including, but not limited to:
-
Some of the interest type descriptions are prescriptive to the point that they fail to represent interests across jurisdictions. For example, the codelist description of voting rights defines this as 'the right of shareholders to vote' whereas in some jurisdictions voting rights may be given to non-shareholders (for example, in companies limited by guarantee in the UK). There is a related question of how different classes of share should be modelled in BODS. For example, preference stock (no voting rights) vs common stock (includes voting rights).
-
The codelist contains ~20 interest types and risks confusing implementers by presenting roles (e.g. board chair) alongside mechanisms (e.g. enjoyment and use of assets) when these might overlap. A reorganization of interest types to reflect the broader ways in which interests can be held through roles and mechanisms may enable more intuitive and comprehensive mapping to real-world interest types.
What user needs are met by introducing or developing this feature in BODS?
The key users who are likely to need this feature are implementers from different countries, who want to use BODS to accurately represent ownership or control interests in their jurisdiction. Again, meeting user needs across jurisdictions is a challenge, as beneficial ownership may be exerted in a broad range of ways within and between countries. User stories include:
-
As an implementer, I want all interest types in my jurisdiction to be representable in BODS, so that I can accurately portray how beneficial ownership occurs.
-
As an implementer, I want BODS interest type descriptions to be sufficiently clear and unambiguous so that I can easily map interest types from my jurisdiction to BODS interest types.
-
As an implementer, I want BODS interest types to be clearly organised so that I can easily find which interests are relevant to my jurisdiction.
There are also data use stories that illuminate some of the modelling challenges.
-
As a civil society activist, I want to know how extensively mechanism X is being used to exert influence over companies in sector Y so that we can provide evidence of how these companies are controlled.
-
As an academic researcher, I want to understand the variety of contractual arrangements that underpin beneficial ownership interests globally so that I can understand different jurisdictions' roles in money laundering cases.
What impact would not meeting these needs have?
-
Implementers may use interest types that do not accurately describe the nature of ownership or control interests in their jurisdiction.
-
Implementers may class large numbers of interests as 'otherInfluenceOrControl', which reduces the richness of data for users.
-
Implementers may use non-standard data to represent interest types, or not use BODS at all.
-
Those investigating BODS datasets may find that the information carried about interests is at the wrong level of detail for them, or that categories are not being used in a reliable way across jurisdictions.
How urgent is it to meet the above needs?
The availability of 'unknown interests' and the fact that most common interest types in most jurisdictions can be represented using BODS mitigates the urgency of this feature, but it should be addressed for BODS v1.0.
Are there any obvious problems, dependencies or challenges that any proposal to develop this feature would need to address?
- Restructuring or adding hierarchies to interest types, if not done well, runs the risk of complicating the standard further and exacerbating the problem it aims to solve.
Feature work tracking
[Link to proposals, bugs and issues in the repository to help track work on this feature]
Revisiting the complete list of interest types currently in BODS 0.3 has left me with some reflections that might help us address this feature in the future.
Conceptual model
Beneficial ownership builds on the concept of legal ownership. Legal ownership in many countries consists of three types of right over an asset:
- rights to use it
- rights to income/profit from from it
- rights to dispose of it
Shareholdings, directorships, trusteeships, appointment of board and other interests can be seen as a breaking down and re-bundling of the three types of right above. There are, though, other interests in entities which are not an aspect of ownership. Other interests like:
- Being a creditor
- Being a grantor / settlor
- Holding share options
- Control via legislation
Getting the right level of abstraction away from detail
The complete and detailed picture of how rights over assets are conferred and managed in capitalist systems is complex, fascinating and ever-changing. BODS is not a standard for representing that complex picture: it is an aid to transparency.
So we'll need to abstract away from the detail while covering the fullest scope of interests. We certainly need to consider representing rights vs representing roles #327.
Getting away from 'ownership and control' language
Along the way, I suspect that our previous language of 'ownership and/or control' will cease to work for us. Already in BODS we're moving in v0.4 to speak of Relationship Statements (instead of Ownership-or-control Statements).
For some of the latest Open Ownership thinking on direct/indirect interests, see key sections of 'Solving the international information puzzle of beneficial ownership transparency'
Setting aside the need to improve the definition of each codelist entry, I see broadly three approaches that could be taken to refine the interests codelist and approach, and one question of implementation guidance.
We'll be doing more research to explore these routes soon, but sharing for early comments or reflections.
Approach 1. Exhaustive codelist
Carrying out detailed research into a full range of interests specified as subject to disclosure in different regimes in order to cross-walk these to common codes where possible, but to create new codes wherever required to capture nuance in different definitions.
Approach 2. Hierarchical codelist
Updating the codelist to be a structured taxonomy, potentially based on top-level categories of:
- Ownership
- Control
- Benefit
- Other
Approach 3. Binary classifications
De-composing the codelist options into a set of different questions about the kinds of interest they convey and the mechanisms by which they are excercised.
E.g.
Interest
- (Shareholder)
- Right to profits
- Right to direct decisions
- Enjoyment of benefit
Mechanism
- Ownership
- Employment
- Statute or regulation
- Private legal arrangement
Modelling trade offs
Reviewing the UK PSC legislation, and the ‘Control of trust’ or ‘Control of firm’ categories (Parts 5 and 6) raises a question for me of when declarations made as a ‘flat’ interest statement should in fact be modelled in BODS as a chain of ownership, or when the standard should have more fidelity to the way disclosure was made.
I.e. As I read it, A may declare that they own company B via a Trust, without disclosing the details of that trust.
In BODS this could be modelled as:
A [individual] -(has an interest in)→ C [unknown trust ] → (has an interest in) → B [company],
however, this would involve the publisher of data converting a single statement of interest as made to them (A [individual] - (has an indirect interest in via a Trust in) → B [company]) into two statements as above, which we may have discovered in practice is harder for publishers to manage?
Put another way:
Real world disclosure rules, and the categories of interest declared through them, may mix things that, in BODS terms should be mapped to simple structure + codelist, and things that can technically be mapped to complex structures / chain of statements.
In mapping between each jurisdiction’s lists of interest, is the best path:
-
(1) Providing mappings that expand out implied ownership chains with unknown entities where required. I.e. if an declared interest type in Jurisdiction X logically implies a trust between person and entity, create extra statements to show this.
-
(2) Allowing that there should always be a simple representation of the declaration (i.e. even if an interest type implies an ownership chain, allow this to be modelled through a single set of entity, person and interest statements, and a relevant code that describes the interest (e.g. controlViaTrust)) and leaving it up to consuming systems to expand out any implied entities if required.
This is example of the use case I use to test complicated cross border ownership/control case involving private companies under control of individuals and Malaysian government owned companies purchasing properties in Australia in a report prepared by EY
https://www.smh.com.au/interactive/2020/_modules/tearout/mara/MARA%20Final%20Draft%20Report%20on%20Aust%20Properties%20140619.pdf
This is a common challenge in Malaysia for reporting and mapping out actual control of private companies, that are ultimately owned by a government entity.
Additional the ultimate government entity usually is headed by a real person (Minister) who by law has control over the selection of the board and decisions.
In our use case for Malaysia, a long list of interest type definitions by BODS is helpful in having some sort authoritative reference/standard and guidance for civil society and journalists in determining whether a government entity has actual control over subsidiaries of private companies, and who is the ultimate person in control of that government entity.
Thanks @kaerumy for insights there.
Looking at Approach 1 (and possibly also option 2), I would note that, for this to work effectively, the codelist may need to be open ended (i.e. regularly updated outside the cycle of standard versioning).
Reviewing how the SC-DEX project call out to Wikirate's relationship types and recalling WikiData approaches to capturing relationships, we may want to consider potential wiki-like infrastructures to support Approach 1 and Approach 2 codelists.
Open Ownership have just published this research piece from @timgdavies and Peter Low. They looked at legislation from a range of jurisdictions to see what categories of interest were mentioned in definitions of 'beneficial ownership'. The categories were diverse, suggesting that standardisation around this facet of beneficial ownership data is a challenge.
Tim and I discussed the implications for modelling choices, and those are included in the paper too.
It would be great to hear from anyone who's thinking about data-handling and standardisation around interest types, and to know whether the findings and ideas in the paper resonate with you.
(And thanks to Tim and Pete for the work on this!)