Aid Fishing in the Yemen: Tracking Saudi Development Assistance to Its Southern Neighbor

AidData tracks development resource flows from a wide variety of non-DAC sources. This continues our blog series focused on "emerging" and non-DAC actors in the development field. Past posts can be found here. This is also the inaugural post regarding AidData's "media-based data collection" (MBDC) efforts.
Many donors not reporting to the OECD Development Assistance Committee (“non-DAC”) lack either the capacity or the political will to publish detailed information about their development cooperation activities. In order to provide a more comprehensive picture of these development finance activities, a small team of AidData researchers recently began developing, testing, refining, and documenting a set of data collection procedures that extract and codify project-level information from news reports. This summer, we helped pilot this methodology by testing its usefulness in tracking Saudi Arabian aid flows to Yemen, which are not published through official channels.
In recent years, the Kingdom has poured billions of dollars into its smaller, weaker, southern neighbor; analysts estimate that Saudi aid to Yemen ranges between $1-2 billion a year. (See recent reporting from Reuters and the Financial Times here and here.) However, details about the nature of these financial transfers are few and far between. Using media-based methods, we attempted to generate the most detailed, project-level summary of Saudi aid to Yemen to date.
Using Arabic and English news sources, we gathered detailed information on some 45 Saudi loans, grants, and credits to Yemen from 2006 to 2010 worth approximately $1.56 to $1.83 billion. These financial flows supported the following sectors: transportation (23-27%), economic development (9-18%), energy (14-16%), water and sanitation (13-15%), health (10-12%), emergency support (5-6%), and education (3-4%). Most of this funding to Yemen came in the form of grants (65%); the remainder consisted of loans (17%) and import/export credits (12%). 
The data show that, during the period in question, Saudi Arabia provided far more than any single DAC country and nearly as much as the top three DAC donors combined. Additionally, the data suggest that Saudi Arabia favors fewer, larger, infrastructure projects—a significant departure from the trend among DAC countries to support smaller, but more numerous, projects in a variety of sectors. 
Overall, we consider the pilot a success, but we did encounter several difficulties, which reinforce the fundamental point that media-based methods are no substitute for comprehensive official records:
     -  First, our use of ranges in estimating total Saudi commitments to Yemen reflects uncertainty about 
         whether some projects should be considered official development assistance (ODA). 
     -  Second, our data collection efforts suggest that Yemen has received, on average, a little over $300 
         million a year from Saudi Arabia—a large sum, but still only a fraction of the estimated $1-2 
         billion that it reportedly receives each year. 
     -  Third, media-based methods are not useful for documenting financial transfers that governments 
         are actively seeking to conceal. Sarah Phillips of the University of Sydney claims that "[i[n 2009, 
         the Kingdom made a direct payment of $US 2.2 billion to President Saleh." Citing unnamed 
         sources from the local donor community, she also reports that Saudi Arabia made a direct 
         contribution to Yemen's Central Bank in August 2010 worth somewhere between $1 billion and 
         2 billion. 
All of these caveats notwithstanding, the results of the pilot provide a rare window into the world of Saudi development finance. They also point to the woeful inadequacy of existing aid information systems for tracking non-DAC sources of development finance. 
Keep a close eye on The First Tranche in the coming months for more details about AidData's forthcoming media-based methodology. We plan to release a codebook and a dataset for a high profile supplier of non-DAC development finance later in the fall. 
James Juchau is an AidData Research Assistant at the Brigham Young University. Brad Parks is the Co-Executive Director of AidData and Research Faculty at the College of William & Mary.   

***Editor's Note***

It was brought to our attention that the chart above is difficult to read. Big kudos and thanks to Tariq (@tkb) for going out of his way to send us an updated, and even fancier, chart for the post.
Yemen Aid

Check out the full exchange here:

Which chart do you think is easier to read?
Tags: south-south cooperationmedia-based data collection (MBDC)


Fascinating post and very promising approach. One important caveat: if the data source is news reporting, it stands to reason that large projects will be disproportionately likely to be found. <br /><br />Before concluding that this pattern is at odds with DAC patterns, I would like to see what the same approach produces by way of data for a DAC donor. Perhaps you've done this already; if so, I would love to read more about it.

This is extremely interesting. A few questions:<br /><br />a. Does the aid data indicate the geographic location where aid has been used in the sectors identified?<br /><br />b. Have you carried out, or do you intend to carry out similar analysis of any Iranian aid to the country? That would round off the political economy considerations of aid to Yemen. Chances are that the Iranian 'aid' may not be official aid at all, but would be interesting to see if there are any stats on this.

@Maurits-<br /><br />Although the literature on Arab aid giving is limited, it is reasonably well established that Arab countries are more inclined to finance large, physical infrastructure projects than their DAC counterparts (see Neumeyer 2004, Shushan and Marcoux 2011, World Bank 2010). Our findings from the Saudi-Yemen pilot project confirm these previous findings. Thus, in a way, this methodology seems particularly well-suited for emerging donors that mostly finance large projects. However, it is also worth pointing out that the application of a media-based data collection methodology revealed a number of smaller projects worth millions or even hundreds of thousands of dollars. Media-based data collection methods are most likely less useful for identifying smaller, less-media worthy projects, but in a field where existing data are either non-existent or extremely incomplete, any significant increase in data collection will likely improve understanding of Arab aid distribution and impact. To put things in perspective, our pilot project captured roughly 20% of the total estimated Saudi aid to Yemen, a far higher percentage than any other project-level breakdown of this particular aid relationship.<br /> <br />The idea of applying the methodology to a DAC donor is a good one. Presumably, if we selected a DAC donor that regularly reports official project-level commitment and disbursement data, the application of media-based methods would help us gauge whether this approach is in fact biased towards larger projects. It might also tell us something about the usefulness of the methodology for generating reliable aggregate aid estimates for non-DAC donors. It could also help us assess identify other sources of bias introduced by media-based methods. After all, you can't correct for sources of bias until you know that they exist! <br /><br />Any suggestions re: which DAC donor to pursue? It seems like it might make sense to hold the time frame and recipient constant in case recipient needs or characteristics are driving Saudi (and DAC) donor behavior.

Christian,<br /><br />Thanks for the thoughtful response. I completely agree that you have an impressive capture rate. Nonetheless, applying your method to a donor whose actual aid projects are known will get you some very valuable additional data: an estimate of roughly what proportion of different type- and size projects you are capturing.<br /><br />For example, you might find that you are capturing 20% of large projects and 20% of small projects for the known donor. But you might also find that you are capturing 30% of large projects and only 10% of small projects. <br /><br />Both of these are quite possible, and both are compatible with the facts you highlight (Arab countries are relatively more likely to fund large infrastructure projects, and you are already capturing some small projects). But they would have important implications for how you might extrapolate from your findings to hypothesize what the uncaptured aid flows look like.<br /><br />In terms of which DAC donor to pursue, I think the UK would be a good candidate in light of its historical connection with Aden. Plus, one suspects UK aid is more likely to appear in English-language sources than, say, French aid.<br />

Thank you very much for posting this interesting analysis. Like you, I have been using news sources in English and Arabic to calculate Arab aid to Yemen and to other countries in the region. I thereby came across two additional problems that are caused by using those sources which you did not mention and which I therefore like to highlight. <br /><br />Problem #1 is that of “multiple announcements”: Often national news agencies report on allocations of money to concrete projects. These allocations are simply the next step in the pledging process, but often years lie between the pledging and the allocation. This creates the impression that in addition to the amount already pledged additional money is being provided. This is usually not the case because when you add up the sums mentioned in the news, you find that they add up to the amount of the pledge. Thus no new money is actually provided, and the amount mentioned in the announcement of an allocation should not go into the calculation. Unfortunately, this link is usually not made by the journalists and news agencies reporting on an aid project. The only way to find out if the assumption of a multiple posting of an amount is correct is through interviews with the national funds of the donor country or with ministries in the beneficiary country. Such interviews, however, are not easy to arrange...<br /><br />Problem #2 is that of actual disbursement. In Yemen’s case, not the whole amount pledged in 2006 had been disbursed by the time the next donor conference took place in 2012. Several Gulf states included portions of their non-disbursed pledges of 2006 into their 2012 pledges. One needs to be aware of this when one uses news announcements of pledges as one source to calculate the rate of assistance by given donor. In this case, the pledges will always be higher than the actual disbursement, even of a donor claims a disbursement rate of 100 percent. Given that you calculate Saudi aid to Yemen for the years 2006-2010, the 2012 pledges will not interfere with your analysis. If you decide to broaden the timeframe, you would need to consider this challenge, be it with regard to Saudi pledges made before 2006 or those of 2012.

Thanks for the awesome work. Do you have detailed information on the amount of aid received for each year and the sector in which it was used?