The Benefits (and Drawbacks) of Coding Aid Projects By Activity

This post first appeared as a comment on the Center for Global Development’s Global Health Policy Blog, and is re-published here.
 
Thank you so much for the post. Extending the QuODA methodology to specific sectors such as health is a valuable step toward better aid outcomes.

 

 
One issue you raise is that “CRS purpose codes remain too broad to track expenditures into most specific diseases or areas.” For some time, AidData has been working on collecting the most detailed and comprehensive project-level data possible. Through the AidData activity coding scheme (which builds on the CRS scheme), we assign aid projects specific codes based on project descriptions provided by donor agencies.

 

 
In this regard, we have already made significant progress: virtually 100% of the multilateral and non-DAC bilateral aid records in the AidData.org portal have been double-blind activity coded. Additionally, AidData is currently double-blind activity coding all 800,000 historical CRS project records, a process which is already 5/6 complete.  (By double-blind activity coding, we mean that two different people individually assign codes to projects, and if their codes don’t match, a third person reviews both and selects the more accurate code). We hope to arbitrate and publish the two rounds of activity coding by 2013. 

 

 
This represents an important step forward in aid transparency. Anyone visiting Aiddata.org can now identify multilateral and non-DAC development projects not only by their overall/dominant purpose, but also by their specific components/activities. Such data granularity allows for more detailed and nuanced analysis of aid allocation.  For example, there are 75 unique AidData activity codes explicitly related to health. If a researcher wanted to identify all reported multilateral and non-DAC projects which provide, say, dental assistance, he can find it by searching the AidData portal.

 

 
To examine the impact of coding projects by activity, consider this $12 million dollar loan for “Poverty Reduction” from the OPEC Fund for International Development (OFID) to Tanzania.  The project focuses on “small-scale, demand-driven community development projects”, including, among other things, “the construction of 14 clinics.” Because the project’s focus is so broad, it would most likely receive a CRS multi-sector code. But with the independently assigned AidData activity codes, the project’s health activities are identifiable through an AidData.org query.

 

 
Of course, there are limits to the versatility of independent activity coding.  Working with project descriptions donors make available, we cannot assign specific dollar amounts to each activity within a project. This means that adding up the value of all projects identified by a certain activity code could be misleading (in the example above, for instance, we don’t know how much of the Poverty Reduction project funding went toward health clinics). Also, if one does multiple searches for different activity codes, adding up the commitments for the resulting projects could result in double-counting (in the event that the same project has multiple activity codes and is identified in more than one search). Finally, the project records cannot “speak to each other”—that is, link to related projects which might relay relevant information.

 

 
Sector-level analysis raises an important taxonomical issue: how many different activities can one “aid project” execute? Donors may choose to define a project as one transaction for a single activity, or as a bundle of transactions for distinct yet connected activities. While AidData prioritizes publishing the fuller project descriptions found in annual donor reports, some donors break down their projects into separate transactions for each activity when reporting to the CRS. Therefore, what appears in the AidData portal as a single project with many activities of unspecified cost, might appear as several unitary projects of specified cost in CRS QWIDS.

 

 
In the future, donors reporting in the IATI format will be able to link project records to related projects--or, even better, full-length project documentation. Until then, researchers interested in sector-level aid allocation can identify the universe of non-DAC and multilateral project records which have been assigned AidData activity codes.

 

 
I share your conviction that “better health aid data will lead to better aid, and hopefully better results.” More robust sector-level data will require more detailed project-level data. This is not an unrealistic goal. Indeed, IATI reporting standards, CRS purpose codes, AidData activity coding, and Open Aid Partnership geocoding all demonstrate that transparent, accessible, and detailed aid information is well within reach. 

 

For more information on the available data resources covering development assistance for health (DAH), I encourage researchers to review this recent Health Policy and Planning article
 
Brian O’Donnell is an AidData Post-Baccalaureate Fellow at the College of William & Mary.

 

 
Tags: iatibehind the curtainhealthgeo-referencing