Aid Type

[andylolz]

Description

The type of aid shows whether the activity is classed as budget support, a project, technical assistance, debt relief, administrative costs, and so on.

This can be either at the activity or transaction level, and is only expected if the activity is in the implementation, completion or post-completion phase.

Proposed test

Firstly:

For each current activity,
if `activity-status/@code` is one of (2, 3, 4)
then `default-aid-type` should be present
or `transaction/aid-type` should be present

Secondly:

For each current activity,
if `activity-status/@code` is one of (2, 3, 4)
then at least one `default-aid-type/@code` should be on the AidType codelist
or at least one `transaction/aid-type/@code` should be on the AidType codelist

[YohannaLoucheur]

Those using IATI data find that when published, Type of Aid is often erroneous. This decreases the overall usefulness of the data, given that Type of Aid is a powerful variable to analyze development activities.

An automatic test could be designed to identify at least some of the mistakes, by combining AidType with Implementing Organization Type. For instance, AidType A01 and A02 require a Government OrgType. AidType B02 requires an OrgType Multilateral, while AidType B03 can only have OrgType Multilateral or INGO (which are easily confirmed via their Channel Code). Perhaps the results of this test could be published as additional feedback to the publisher, and added to the Index score in the medium term.

[andylolz]

Really interesting, @YohannaLoucheur. The test you describe sounds like it might be better included in the Standard Ruleset, and tested by @VincentVW’s IATI bugtracker, or as part of the next iteration of the IATI Validator.

It sounds like useful feedback for publishers, but ultimately the fixes sound technical and relatively low bar. What do you think?

[YohannaLoucheur]

This is about data quality, which the Index should try to better assess IMO. It's not sufficient to put something in the field, that something should be reasonably accurate. The Index methodology tests the quality for other indicators, I'm proposing a way to do it for this one as well.

Not sure I understand why you see the fixes as being technical? This is bad data captured in publishers' systems (including ours, unfortunately). The test would be a way to signal to publishers that they have a problem, and that it matters, which hopefully would convince them to correct it.

[andylolz]

Well, perhaps I am being overly opinionated here. My apologies – Your point is well taken @YohannaLoucheur, and we’ll certainly discuss it. Thanks!

To check, though: My impression was you were not proposing a change for the next Index, but rather suggesting we provide information on erroneous aid type data as feedback to the publisher initially. Is that correct?

[YohannaLoucheur]

Ideally it would be in the next Index, but I'm being pragmatic - perhaps it would be seen as too drastic a change, not enough time to consult on it, or maybe you'll find the test is not robust enough yet. What I'm suggesting is that if it's not in the next index, there would still be value in running it and providing results, with a view to integrating it fully in a future index.

[VincentVW]

Very interesting data quality issue @YohannaLoucheur ! Sounds useful for at least additional feedback imo.

These functional dependency type of data quality issues are an area where programmers (or me at least) often are not aware of while they might be obvious when you know the context better. So if there's any other frustrations when using the data where the validity of one element (aid type) is dependent on another element (organisation role+type) I'd be happy to hear.

Thanks for the mention @andylolz , I'll put it on my list to check in OIPA and show on IBT. Just realized there's actually people using it, might as well finish it :)

[YohannaLoucheur]

Thanks @VincentVW , will do! And always happy myself to provide feedback and help figure these things out. I have 0 programming expertise, but I do know the data and (some of) its potential uses. I think we all need to work toward enabling users to make full use of the existing data/standard, rather than adding more elements or attributes to the standard.

[andylolz]

Thanks for the comments here.

These tests will remain as proposed.