New methodology for identifying government entities

The proposed approach has a high risk of different publishers ending up with different org IDs for the same government entities (as it is already often the case in IATI for other types of participating org.). "To use existing public information for each country, either its Chart of Accounts or the national budget" is not a very straightforward thing to do. It can be subject to interpretations and changes over time. In addition, it is a very time-consuming process. From this point of view the list of codes for 50 countries is very helpful. It would be good to extend it to other countries, so that it can be used as the "source of truth" by all publishers. The possibility for publishers to propose a code for those government entities that are not part of the list would also be helpful to build a solid source of information over time.

Evgenia Tyurina, many thanks for your comment, I really appreciate your engagement on this! And really welcome further questions and comments from everyone on this.

Just wanted to pick up on a few of your points:

1. Ease of implementation: indeed,   we went through budgets and did this for an initial set of countries in order to make life easier for everyone. It's not a particularly onerous job, but of course it makes much more sense for this to be done once per country rather than many times by every publisher. But I think it's important to emphasise -- the authoritative codes remain those in the government budget documents. We're not trying to establish a new codelist here -- we're just pulling out these codes as a courtesy, to make them much easier to use.
2. Stability: while Charts of Accounts do frequently change, the administrative classifications at the highest level (i.e. Ministry), which is what we're normally interested in, are very stable over time. While new codes do occasionally get created (e.g. when a new Ministry is created), they don’t get re-used (with incredibly rare exceptions). Occasionally (every five years or so, or whenever there is demand), the codes that have been collected could be updated.
3. Updating: it's not a particularly onerous job, but of course it makes much more sense for this to be done once per country rather than many times by every publisher. It would be great to establish a process to add new countries over time  .

With this approach, I think it is possible that publishers will at some point use a different or an outdated code (e.g. referring "Ministry of Finance" rather than "Ministry of Finance and Planning" after Ministries have been merged). But that is far more useful than using nothing at all, as it's much easier to map recognisable codes than to have to code from scratch. There will be cases where it doesn't work, but after ten years of searching for a solution on this, let's not let the perfect be the enemy of the good!

Evgenia Tyurina - a tweak that could help this issue (unclear to me if it would break something else) would be to move the source information which identifies the year that the codes are taken from into the code e.g. LR-COA-301 would become LR-CoA-2021-301......that way if you had data for e.g. 2005, you could select the appropriate codelist for year the data is from. LR-CoA-2021-301 is a different code from LR-CoA-2005-301 so no problem if the organisation changed. It would mean that users need to use the name narrative element to combine orgs that had not changed between the two different CoAs but that seems possible to do, whereas the issue you describe would be very difficult to solve otherwise. Users who did not want this effort could just pick the a single codelist for all their data irrespective of date and not be worse off than before.

EDIT: in case of misunderstanding, despite suggesting an alternative approach with different costs/benefits, I think having this new tool and setup plus the underlying work to gather all the codes reflects great progress on a long-standing issue!

I think this problem also exists in other IATI codelists e.g. I believe FCDO now use the code GB-GOV-1 code formerly used by DFID? If you were using IATI data to do a study to compare the allocation choices of aid for donor agencies that were independent from foreign ministries from those that were combined, the code re-use would mean that your results would be invalid, unless you were manually adjusting the years.

I think a lot of these choices reflect the IATI decision not to prioritise use of data that is a few years old - but that should be clearer to everyone and maybe re-discussed!

Hi Matt,

We considered including the year in the identifier but for several reasons don't think this is a good idea:

1. In order for this methodology to be used, it needs to be as simple as possible for publishers to know which code they should be using to refer to a particular entity. So if we don’t really need to include the year, then we should avoid doing so. A 90% perfect solution that gets used is much better than a 100% perfect solution that doesn’t get used.
2. From reviewing a number of budget documents, it appears that while new codes do occasionally get created, they don’t get re-used. Occasionally (every five years or so, or whenever there is demand), the codes that have been collected could be updated.
3. There is also an enormous amount of consistency in the use of codes over time. For example, for Peru, the following Ministries existed in a 2004-2008 BOOST dataset versus a 2014-2016 BOOST dataset (source):
   
   Only those Ministries highlighted in yellow changed over a ten year period. Two were entirely new creations; the ones that changed are recognisably very similar Ministries.
4. On the use side, if you included the year in the identifier, you would either need a mapping file to map between codes from different years, or alternatively handle these identifiers differently (by generally ignoring the year, so seeing that LR-COA-2021-301 is almost always the same as LR-COA-2005-301).
5. Maintaining multiple lists for each country plus mapping files would create significant additional burden and complexity, and again, given that these codes rarely change at the top level, I think there would be limited benefit.

I agree that there's a broader issue that we should think about in terms of "decay" of codes over time. For example, where a publisher uses a CRS Sector code that used to exist, but has now been withdrawn, and should be updated to use a new Sector code. I think that merits a separate discussion as it is much broader than this issue.

A quick update from me: we've made some more improvements to Gov Org ID Finder so you can now more easily browse and download codes for 53 countries. For example, see the country page here for Liberia:

https://gov-id-finder.codeforiati.org/countries/LR

If you'd like to contribute codes for another country, or if you want to propose changes to the codes that we've currently captured, please get in touch!

The COA prefixes are now available on org-id.guide for every country, so using a code with a COA prefix (e.g. LR-COA-310) will now validate. The Code for IATI OrganisationRegistrationAgency codelist has also been updated to reflect these changes. It is automatically updated every day to keep in sync with the org-id.guide lists.

Again, please do get in touch if you would like to contribute codes for another country, or if you'd like to propose changes to the codes that we've already captured.

I'm late to the party but what is COA?  Why would we use GOV as some publishers already do and I believe is the guidance in the Org ID guide? 

Hi Michelle Levesque  thanks for the question - COA stands for Chart of Accounts. This is the set of classifications that each government uses as part of its budget and financial management processes. IATI's guidance on organisation identifiers is to use a prefix which is on the OrganisationRegistrationAgency codelist (which is maintained by Org-ID.guide), COA prefixes are now available on this codelist for every country.

It's a good question as to why we didn't opt to use the "GOV" prefix. The answer is a bit complicated, given the ways that the "GOV" prefix is currently used.

There are two entries under on the OrganisationRegistrationAgency codelist for the prefix "GOV", for Canada and the UK (searchable version here - search for "-GOV"). Among reporting organisations, Korea and USAID are also using the prefix "GOV", but there is no corresponding entry on the OrganisationRegistrationAgency codelist.

In the case of the UK, this is a set of codes used by the UK to refer to UK government IATI publishers, basically following the format of the (now deprecated) IATI OrganisationIdentifier codelist. However, the "GB-GOV" codelist was last updated in 2019, and is currently inaccessible.

In the case of Canada, the CA-GOV codelist is actually extracted from the Canadian Chart of Accounts. This codelist appears to be regularly updated. That's consistent with experience which suggests that COA codelists (even if they are not always machine-readable) are very likely to be maintained and kept up to date, as they are necessary for governments as part of their core financial management processes.

In the case of Canada, CA-GOV and CA-COA would presumably be the same (I haven't looked at this case in detail - Yohanna Loucheur ?). In all other cases, we opted to create a new codelist for every country (except for Afghanistan, where the prefix AF-COA already existed, and happens to follow the same methodology). We didn't want to use a namespace that was already used in a different way, in order to avoid confusion or conflicts. We wanted to make a standardised methodology that works for every country, and can be used straightaway, especially when referring to governments as implementing organisations. "GOV" is currently mostly used by some publishers to identify themselves, and generally in a less systematic or consistent way than the COA approach, so it's not easily scalable to all other countries. Though I have to admit, CA-GOV-... does look nicer and more authoritative than CA-COA-...

I understand IOM doesn't publish the names of your implementing organisations. But in case you are working with governments and wanted to state that, I would be very interested to know if you think the methodology (summarised on Gov ID Finder) works for you!