Reducing Corruption and Improving Accountability in Aid Projects through Targeting
More precisely targeted projects are on average less likely to suffer from the capture of funds due to corruption or other forms of diversion.
In January 1996, the World Bank approved two transportation infrastructure loans to Kenya. The first, the Urban Transport Infrastructure Project, suffered from significant problems with fraud and corruption that led to three World Bank staff members being fired and 11 companies temporarily being barred from bidding on World Bank projects. The second, the Nairobi-Mombasa Road Rehabilitation Project, was rated highly satisfactory and appears not to have suffered from the same problems with corruption.
In an article forthcoming in International Studies Quarterly, I argue that the key difference between these two projects is the fact that the second project was more precisely targeted. Using original data from almost 600 World Bank-funded investment projects, I show that more precisely targeted projects are on average less likely to suffer from the capture of funds due to corruption or other forms of diversion.
There are three reasons why international development agencies might want to make projects more precisely targeted in order to reduce the likelihood of international funds being captured. First, the accountability relationships in delimited projects are likely to be clearer, such that bureaucrats or government officials responsible for a project can be identified and sanctioned for poor performance if necessary. Second, outcomes are more easily observable in more delimited projects, such that it is more evident if such sanctioning is necessary. Third, the stakeholders themselves are more easily identifiable in more delimited projects, such that the hurdles to collective action and group sanctioning behavior are lower.
For the projects in the dataset, I coded their level of targeting and whether or not they suffered from capture. (The complete coding rules for both the outcome and the explanatory variables are available in the replication materials for the study.) Classifying World Bank investment projects into nine targeting categories, the data reveal that projects targeted at a single region or a single city, for instance, experience lower rates of capture than those targeted at multiple regions or multiple cities (Figure 1). In the paper, I use a series of statistical regression models to show that this relationship holds even when controlling for the total size of the project and the specific sector and theme of the project. Indeed, the relationship holds even when country fixed effects are included in the model, implying that a more precisely targeted project in a given country is less likely to suffer from capture than a more diffusely targeted project in that same country.
The outcome variable is constructed from information contained in the World Bank’s Implementation Completion Reports. When those reports mention corruption directly or describe problems with the financial management, procurement practices or audits in the project or describe political interference in allocation decisions, I code the project as suffering from capture. The level of targeting is likewise coded from World Bank documentation.
This coding scheme is undoubtedly inexact and likely to generate both false negatives and false positives across the sample of World Bank projects that have been coded. In the article, I provide prima facie evidence of the validity of the capture variable by showing that projects are significantly more likely to be coded as suffering from capture in countries that are regarded as more corrupt by the Worldwide Governance Indicator’s Control of Corruption measure or Transparency International’s Corruption Perceptions Index. In addition, insofar as World Bank staff may have certain incentives to misreport corruption in the Implementation Completion Reports, it is hard to think of a reason why they would be biased toward doing so in precisely-targeted projects, such that the main empirical results of the article should not be biased by noise in the data.
The policy recommendation that comes out of the research is that development agencies should use more precise targeting in order to create better accountability relations in the projects that they finance. My previous research suggests that this may already be happening at the World Bank: I find that World Bank projects are less likely to be nationwide projects in more poorly governed countries. Insofar as targeting reduces the amount of corruption in development projects, this should imply that international aid dollars are being better used, delivering more goods and services to impoverished populations.