Working with IATI-data — Lessons learned
Last week we launched a connection between our (beta) service and IATI. It enables users to analyse aid activities by just inputting an IATI identifier. Our service connects to IATI, pick out the relevant financial data and returs an analysis. The IATI integration is an early beta, but we are really happy about it and have already received a lot of comments and feedback (Thanks!).
In this post, we will be reflecting a bit on how it is to work with IATI-data. Hopefully it can be useful both to others thinking about how to use the data, and to those working with/reporting data. We try to be constructive, and we end off with a suggestion that we think would be really interesting!
A quick note: a technical overview will be posted shortly.
A starting point for our comments is that we have a very specific use of the data. We aim to analyse financial aid flow, to identify how those flows are affected by such things as exchange rate fluctuations. Volatility can affect the predictability of aid, one important factor affecting effectiveness. We also look at transaction outcomes, among other things.
Funding most often flow from donor, through one or more implementing organisations to the intended results or recipients. To understand the full efficiency of such a flow requires good data throughout the chain, from the end point back to the donor.
This has been our main interest when integrating data from IATI. And also our main challenge.
- There is varying quality of data when it comes to the connection from one activity to different IATI-activities and organisations. The “delivery-chain” very often has broken links, making more advanced analyses impossible. We understand we are not the only ones interested in this, and think the standard as such would benefit from a higher degree of minimal traceability.
- Trying to establish chains, tracing flows, we often have to rely on organisational roles to establish who is who. It’s the starting point in trying to provide an analysis of who’s affected by exchange or inflation risk, for example. Few donors take on risk exposure, so it’s highly relevant to ensure a report focuses on implementing organisations instead. We found that there is not a strong enough common use of roles, making it harder to establish a clear chain.
We have realised that we are in no way the only ones interested in the question of traceability, and we have looked into some very interesting discussions on the topic (recommended!).
A part from the minimal tracebility of having clearly connected activities, an important question for us is the financial data quality.
While we understand that IATI is not a accounting system, we would still argue that the quality of financial data is a very important question. Trying to use IATI-data as a foundation for practical accountability or planning is hard if the financial data is not good, or clear, enough.
An example from our use of the data:
- We are really interested in how exchange rate fluctuations have impacted funding or cost (or can in the future). Understanding this builds on data about involved currencies and the dates of commitments and transactions. But this data is very often lacking or of bad quality. It turns into a puzzle of assumptions, where we have turned to project /donor countries etc. to fill the gaps.
It boils down to a case of the better the data, the better the possible analysis. But we would like to highlight that this can be seen as a question of usability and perspective, not only limited to our perhaps narrow service.
Not having good enough data on currencies and disbursement dates makes the use of the data for financial accountability very shaky. Exchange rate can easily affect the value of received funds with 40 percent (+-20) or mor in one year.
The value likely most important in recipient countries is the value received, not just the donor disbursement in the original currency.
Take, as an example, Swedish IATI-data (which we have worked a lot with). The financial activity is reported as a yearly aggregate, always set to a end-of-December date. The amounts are reported not in the actual SEK used, but using an USD equivalent, calculated using a OECD yearly average rate. Trying to use this data to understand the value of funding received in by partner organisations or countries will have a very high amount of uncertainty. And trying to use the data for future planning is likely as hard.
There are many other interesting discussion around how the data can/should/is intended to be used. We are happy to continue it, at the IATI Discuss or elsewhere!
Edit: Check this article from Development Gateway with a broader analysis!
We would just like to end off with a suggestion however. As there is a lot of great data aggregated in IATI any effort that would increase the traceability through the aid system would be really valuable, in our view. We understand that there are a number of limitations on the current data, and the way changes could be made. So instead of trying that, what about a solution that tries to connect as many activities as possible?
It would be really exiting to see what could be done with machine learning/AI, using the data already at hand.
Not in the current data, of course. But as separate data, that could be used at sites like d-portal or others. A possible result could be a substantial increase of the traceability, adding to the value of the current data. With some work, all activties could be matched, based on a confidence interval.
That is all for now! As always, we are happy to hear your thoughts: reach out on Twitter or send Númi and email! (or in this case, you can also write something on IATI Discuss).
