If you are familiar with the work here at IPD, you know that we love our data. And in the world of development aid data, there is a host of suppliers to love. Everyone from IATI to AMP to AidData is striving towards the same goal: to use transparency and open data to make more efficient aid decisions and allocations. But how do each of these suppliers compare with one another? Is there potential for them to work together and revolutionize development for the better? IPD’s most recent project consists of a deep dive into the world of the World Bank’s BOOST Budget Data and Mapping for Results (M4R) data in an attempt to answer these burning questions.
Launched in 2010, the World Bank’s BOOST initiative encourages governments to publish their budget data on user-friendly platforms. In addition to the BOOST budget data, the Bank provides data on project aid allocations through their Mapping for Results initiative (M4R). M4R not only provides a visualization of the locations of the Bank’s activity, but is also a part of a larger movement aimed at providing access to information regarding funding, poverty indicators, sector (health, agriculture, etc.) allotments, and most importantly, results. The success of the transparency initiatives is contingent upon the degree of their compatibility. Compatible budget and aid data sets allow government expenditures to be easily mapped against project aid, encouraging more efficient allocation of funds and decreasing overlap between government and donor projects.
Our research was two fold. With Kenya as our subject of analysis, we narrowed our scope of comparison to the data sets’ geographic divisions for the country (administrative division precision codes, in technical terms) and the description of the aid project or budget expenditure for the health and agriculture sectors, which are referred to as activity codes.
Two questions drove our analysis:
1) Are the geographic precision codes in the budget data, which indicate the type of location (city, district, etc.), compatible with the aid project data?
2) Do the activity codes, which indicate the types of activities taking place, in the health and agriculture sectors of Kenya match across data sets?
Let’s tackle precision codes first. For Kenya, the administrative division codes were messy. A new constitution adopted in 2010 drastically changed the country’s administrative divisions. While the BOOST data included these newly mapped divisions, the M4R data sets used pre-2011 divisions. The result? An overlapping web of administration division that makes these two data sets extremely difficult to reconcile.
The following three maps illustrate the different ways that Kenya has been divided. The first map shows the provincial system that was in place prior to the 2010 constitution change:
Figure 1: Provinces, Kenya (pre-2013)
The provincial system was replaced by the county system shown in the second map:
Figure 2: Counties, Kenya (post-2013)
The third map shows the local authorities (municipal, county, and city councils). These, along with electoral constituencies (not shown), are other types of administrative divisions that exist concurrently and do not form part of the county system:
Figure 3: Local Authorities, Kenya
When it came to activity codes, the findings were similarly incongruent. After examining all of the potential options for budget expenditure classification and aid activity codes, we attempted to match each code with a potential partner in the accompanying data set. For most cases, in both health and agriculture, there was no natural partner. The table below shows the pairings of categories for the agriculture sector. Green indicates strong compatibility, yellow is moderately compatible, and red indicates incompatibility.
Figure 4: Compatibility of government budget codes and World Bank project activity codes
Our curiosity about whether aid suppliers were on the same trajectory planted the seeds for our research. The difficulty in this type of research, however, is the enormity of the task at hand. As mentioned previously, there are numerous major players in the open data game, and each one has its own system of classifications and categories. Our research only analyzed two data sets provided by one supplier for one country and found limited compatibility. What can we expect, then, as we add more variables to our equation?
One very real concern of our team is that, as we compare different countries using different data suppliers, we will run into even greater inconsistency that could make analysis not only difficult, but also misleading. This prediction, of course, is not meant to discount the potential of open data, but rather to provide insight into making this type of data use even more effective.
Data suppliers must think critically about how they are classifying activities, expenditures, and regions. It is important that the World Bank and other donor organizations work closely with their country partners to develop uniform administrative divisions and activity codes. We believe that an established, smaller number of general activity categories will improve the system. In terms of the BOOST initiative, aid classifications and precision codes may need to be tailored to each country based on national budget expenditures.
As our analytical gears continue turning here on the Open Aid team, we posit this question - how can we improve compatibility and comparability among data sets to make the most of open data initiatives?
Deirdre Appel is a Master’s candidate in Global Policy Studies at the LBJ School of Public Affairs at The University of Texas at Austin.
The issue of climate change is heating up the policy world. This past November we witnessed the landmark joint agreement between the United States and China to cut carbon emissions and increase the use of renewable energy. This December, we expect to see more of the same progressive changes at the Conference of the Parties (COP) climate talk in Paris. But, will it be enough? According to some experts, the reformed emissions targets are still too moderate to save us from irreversible global warming. If we are unable to stop Earth’s atmospheric temperatures from rising 3.6 °C, we may see consequences such as rising sea levels, extreme droughts, food shortages, and uncontrolled disease outbreaks. These dire visions are already a reality in Malawi, where catastrophic flooding has displaced at least 174,000 people since mid-January.
With these challenges in mind, it seems our best bet lies in the adaptation and mitigation of the effects of climate change. Developing nations that will likely suffer the most from climate change are often the least responsible or prepared for its effects. The World Bank, along with many international sponsors, has sought to remedy this unfortunate truth by funding climate-oriented projects in developing countries. The difficulty for many of these large organizations, however, now lies in seeing the overall picture of their many individual projects. Are these donors sending climate-relevant money and projects to the neediest of countries? How do we define this “neediness?” As part of the Climate Change and African Political Stability Project (CCAPS), we attempted to answer these questions by analyzing World Bank projects in Sub-Saharan Africa from 2008 to 2012 for climate change relevance using a coding methodology CCAPS first piloted in 2013 for Malawi. We then mapped these various project locations and funding amounts in an attempt to visualize the relationship between need and total World Bank involvement. Unsurprisingly, we found that there is no definitive way to answer this question. The more we explored, the more questions we found.
To evaluate the efficacy of allocation in World Bank climate funding, the IPD Climate Change Team compared the location of World Bank climate-relevant projects to the CCAPS climate vulnerability index. This vulnerability index is a composite of four baskets of indicators that together predict an area's vulnerability to climate change: physical exposure, community and household resilience, governance and political violence, and population density. This was how we broadly defined the relative “neediness” in each country. Figure 1 shows the result of this exercise:
At first glance, the World Bank’s climate projects appear to be relatively distributed based on vulnerability. Highly vulnerable areas, such as Somalia and South Sudan, have more project locations than the less vulnerable southern states of Namibia, Botswana, and South Africa. However, this map presents an oversimplified analysis. It excludes information on project size and funding amount. It also weights each project location the same, whether the project has one location or 100 locations, and does not consider funding at the national level.
In order to address these oversimplifications, we made a series of maps that explored different ways of analyzing the same data. We first wanted to look at the amount of funding that went to each project. Logically, the most vulnerable areas should have received the most funding. To test this hypothesis, we used different sized dots to represent the amount of funding for a given project and normalized this amount by the number of locations associated with the project. In other words, we took the project’s total budget commitment and divided it by the number of locations. The map below shows this relationship between activity funding and vulnerability.
Although there are only five climate projects with 24 locations in the less-vulnerable, southern states (South Africa, Namibia, Botswana), this map highlights that the projects are much larger than those found in more vulnerable areas such as Somalia or Central African Republic. At $250 million, South Africa had the sixth largest climate project, out of 291 climate projects, in all of Sub-Saharan Africa. Not far behind, Botswana received the eighth largest project at $242 million. To be clear, this map does not consider funding at the national level or population size, and assumes that each location of a project receives the same amount of funding, which is likely inaccurate.
The more we explored the data, the more the information contradicted our previous conclusions. Though each map attempted to answer the relationship between vulnerability and World Bank projects, each map showed a different facet to the same question. The only thing we can conclude is how difficult it is to evaluate the effectiveness of the World Bank’s response to climate change. Aid allocation depends on issues more nebulous than a country’s level of vulnerability.
Forthcoming research will explore “10 Ways to Evaluate the World Bank’s Climate Change Projects in Sub-Saharan Africa” and provide a more detailed comparison of World Bank climate change projects to climate vulnerability.
IPD's Ilse Munoz-Ramirez, Senior studying Plan II and Geography, and Tiffany Wang, Freshman studying Supply Chain Management, co-authored this post.
Maps by Krista Rasmussen, MA candidate, Global Policy Studies, LBJ School of Public Affairs.
During Fall 2014, IPD’s Global Health Team was tasked with working on maternal and child health donor briefs for various countries around the world. I was assigned Niger. Before this project, the only thing I knew about Niger was that it was a country in Africa. To be completely honest, I didn’t even know how to pronounce it correctly. Listen to the correct pronunciation here.
Through my research for the brief, I learned that Niger has a poverty rate of 56 percent, ranking it last among all countries worldwide in the 2012 Human Development Index. According to the World Bank’s 2013 Country Partnership Strategy Report for Niger, the Niger government remains highly committed to achieving the Millennium Development Goals, though progress is slow. One particular area of concern in Niger is its high maternal mortality rate (MMR) (590 per 100,000 live births).
The most recent related project in Niger was funded by the World Bank in 2007, under the auspices of the Bank’s Reproductive, Maternal and Child Health sector for the Multi-Sector Demographic Project. This $10 million government-sector project ended in March 2013 and was successful in meeting its target goals. By the end of the project, one of the three primary goals was exceeded (increased contraceptive use among women aged 20-24), one was met (number of work plans prepared and adopted), and one was missed by 0.7 percentage points (off a targeted 28 percent gain in contraceptive knowledge among women).1
The graph above allows for some further understanding of the situation in Niger. It has one of the highest maternal mortality rates among countries with the highest total fertility rate. That means that it has one of the highest percentages of women dying when they give birth. According to a 2010 report from the World Health Organization, 79 percent of the reason why Niger citizens die is because of the transfer of communicable diseases, the most for any any African country. Additionally, there are only about 0.5 doctors and 3 hospital beds for every 10,000 people in Niger, meaning even if you get sick, there is a huge lack of infrastructure to treat you. It also doesn’t help that 66 percent of the Niger population is living under $1, furthering their lack of access to adequate health care.
In 2010, the World Bank launched its Reproductive Health Action Plan (RHAP) with the primary objective of helping client countries improve reproductive health outcomes, focusing on alleviating poverty and strengthening health systems. Considering Niger’s strong performance on the 2007 World Bank project, its high maternal mortality rate and total fertility rate, the government’s commitment to development, Niger is an ideal candidate for support under the RHAP, which strives to help the higher burdened countries. Why then has it been largely ignored, with just one relatively low funded project throughout the five year Action Plan and no additional funds approved during the Action Plan timeframe? It appears as though the World Bank has missed an opportunity in Niger and unfortunately, many women are losing their lives because of it.
Saumya Wali is pursuing her BA in Public Relations and a Business Foundation Certificate at UT Austin. She has been a member of IPD since Fall 2014.
Open data is a popular tool for researchers, policymakers and donors to study international aid in a new light. Stakeholders can analyze vast streams of detailed data to examine how effectively aid is targeted and implemented. For example, open data gives researchers an unprecedented capacity to track individual development projects on a subnational level from the funding stage through the implementation process. As a student researcher on IPD’s Conflict and Development Team, I utilize open data, geocoding and mapping software to collect, analyze and visualize data on the relationship between violence and international aid.
Our team’s most recent project sheds light on an obstacle that researchers using open data often face in the quest to enhance aid success. In violence-ridden areas, which are often in most need of aid, a lack of sufficient data inhibits accurate analysis of problems and, subsequently, successful aid targeting. How effective can the open data revolution be in areas still plagued by conflict? It seems that, while the use of open data can improve subnational aid allocation, a conflict-driven data void obscures the need for aid in crucial areas. Donors that are serious about development should not allow the wealth of open data to distract them from the importance of conflict resolution.
To enhance the existing literature on the relationship between violence, resource wealth and development, IPD’s Conflict Team is using open data to categorize and analyze mineral resource extraction locations and armed conflict in dozens of countries worldwide. Using subnational data provided by the US Geological Survey (USGS), we geocoded thousands of resource extraction locations in dozens of countries. By merging this data with information on rebel groups and armed conflict locations from the Armed Conflict Location & Event Data Project (ACLED) dataset, we sought to understand the nature of civil conflict onset and duration in resource-rich locations. While we have made progress using this open data, a thorough subnational analysis is hindered by the very conflict events that we seek to understand. Nowhere is this paradox more pronounced than in the case of the Democratic Republic of the Congo (DRC).
So far, we have geocoded resource locations in the DRC from 2003-2010. As we compiled datasets of resources and extraction locations, we assigned each location a precision code. These codes establish how precisely we can locate the given extraction point: 1=mine/production facility; 2=nearby city; 3=district level; 4=province level. We always aim for the most precise location information and consult numerous sources, including media reports and government data, to try and pinpoint the exact location. Yet, despite our best efforts, mines and production facilities in the Sud Kivu province of the DRC have proved impossible to track down. Most receive a precision code of 4, diminishing these data points’ analytical value.
Last year, I came across an article in the Washington Post that shed light on the data void in Sud Kivu. Most of Sud Kivu’s mines are located in the Shabunda territory, which is nearly the size of Belgium and largely controlled by a ruthless militia. In the words of the province Minister of Mines, the government “can’t go there.” Thoroughly isolated in the country’s east, the territory has no telephones, no postal service and no radio. A few landing strips serve as the region’s connection to the outside world. While Shabunda is rich in natural resources, extraction is dominated by the Great Lakes Mining Company, an operation run by a rebel group. The rebels who exploit the resources in Shabunda benefit from the territory’s remoteness and lack of development; they stockpile riches in secret while the government and external actors stay away. Due to the volatility in the area, only one international NGO, Médecins Sans Frontiéres, still operates. Yet, astonishingly, Shabunda is home to more than one million Congolese.
These million people face devastating levels of hunger and malnutrition, as rebel violence has destroyed agriculture, fisheries and livestock, which in turn has destroyed jobs and livelihoods. Measles and smallpox have reappeared. Widespread rape by rebels has led to high rates of HIV infection. Over 95% of children lack access to education. A million people, whose fate is largely in the hands of a few violent rebel militias, are stranded in a dire humanitarian crisis mostly unknown to the outside world.
Of course, many researchers and policymakers are aware that the DRC has long been mired in violent conflict. Yet, for those invested in the conflict’s resolution, a complete lack of data on one of the nation’s worst-suffering territories is an incredible hindrance to progress. This example simply reinforces the initial argument that without sufficient data in conflict areas, targeted assistance cannot be effective. Sadly, collecting sufficient data may be impossible in areas where rebel groups benefit from the cover of endless violence.
Our efforts to geocode and map thousands of resource and conflict event locations have already yielded significant inferences into the relationships between resource wealth and violence. Our data visualizations will provide donors and policymakers with evidence of how rebel activity can stunt development in areas that are resource rich. However, a lack of precise data on one area in the DRC offers an equally significant finding: without conflict resolution in key locations, the open data revolution is incomplete.
Blog post by Anna Scanlon, undergraduate at UT Austin majoring in International Relations and Government.
Map by Amy Leung, who is finishing up her BA in Geography and Sociology at UT Austin.