https://orcid.org/0000-0003-0521-2435
![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
Traditional, hereditary chiefs are an integral part of the development infrastructure in many African countries. To capture chiefs’ behaviour as agents of development and understand the accountability mechanisms they face, we conduct a field experiment with 200 Malawian village chiefs, documenting how they distribute a valuable development good – iron roofing sheets – as we sequentially add monitoring by donors, subjects, and the state. We find evidence that even in the absence of formal accountability institutions, chiefs are responsive to monitoring by all principals. However, principals have competing demands: while most principals prefer allocations based on need as classified by the local community, a subset of the chief’s subjects – his relatives – prefer to receive sheets themselves, regardless of need. When subjects are informed about the availability of sheets, relatives are able to capture allocations, overriding other principals and causing discontent. Altogether, diversion is minimised when chiefs are monitored by the donor, and only the donor. When chiefs are monitored by all their principals simultaneously, diversion is not significantly lower (compared to control), but dissatisfaction among subjects is greater. This study adds to the literature on chieftaincy and highlights the role of common agency in the design and analysis of development interventions.
Disclosure statement
No potential conflict of interest was reported by the authors.
Supplementary Materials
Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2019.1703955
Notes
1. Notable exceptions are Olken (Citation2007), Serra (Citation2012), and Ottone, Ponzano, and Zarri (Citation2014), the latter two of which are lab experiments.
2. As with all studies attempting to determine causal relationships, this article develops a theory of Malawian chiefs that, we argue, may be generally applicable. However, it should be noted there was variation within our sample, and that our arguments and findings should not be treated as definitive evidence regarding the behaviour of all chiefs, either within or outside Malawi.
3. Chiefs in Malawi are part of a seven-level hierarchy that culminates in a handful of paramount chiefs. At the lowest level, the village headmen oversee about 100 households. Village headmen are immediately overseen by group village headmen (GVH). Above the GVH are several levels of Traditional Authorities (TAs), and then the Paramount Chiefs (PCs). At the time of data collection, there are 23,104 village headmen, 3,994 group village headmen, 264 traditional authorities, and 6 paramount chiefs in Malawi.
4. The chiefs in our study are currently paid a monthly stipend of 2500 Malawi Kwacha, or $3.50.
5. Chiefs have no means of coercion and recent literature suggests chiefs even have less control over land than has generally been assumed (Bennett, Ainslie, & Davis, Citation2013).
6. We were constrained to this sample size by our grant budget. Based on pre-treatment power analyses, we expected this sample to generate slightly less than 80 per cent power.
7. Ethical review and approval for this research project were provided by the University of North Carolina, the Pennsylvania State University, and the National Commission for Science and Technology in Malawi. A copy of the application can be provided upon request. We also pre-registered our indicators, most analyses and our (incorrect) hypotheses as EGAP 20160112AA.
8. Tearfund is an international NGO based in the United Kingdom. Representatives from Tearfund Malawi were involved at all stages of the project. None of the respondents in our study had heard of Tearfund, though the appearance of a new NGO in the community is not out of the ordinary.
9. We painted the sheets so we could more easily trace them, and because we were asked to distinguish our project from government-sponsored distribution of iron sheets occurring at the same time. White was one of only a few colours not associated with a political party.
10. Only male village headmen were enrolled. In interviews, we were repeatedly told that female chiefs ‘do not steal.’ We know of no systematic evidence on this claim. Nevertheless, multiple sources were confident that including female chiefs would substantially dampen our treatment effect, so, to maximise statistical power, we avoided sampling in matriarchal areas.
11. For each treatment, the chiefs were ‘treated’ both through the message, which acted as a notification regarding the type(s) of monitoring they would experience, and by experiencing these type(s) of monitoring in reality. These two elements of the treatment are bundled, and we are unable to disentangle them.
12. Discussions with villagers would not have been necessary if, at the time of the site visit, the sheets were already installed, white spirals facing out, exactly where the chief said they would be. However, this did not happen in reality, so discussion with villagers was, in fact, part of this treatment in all cases.
13. The research manager later received several follow-up phone calls from the officials who received the letters, suggesting the letters were read and taken seriously, and completing the letters posed a reasonable threat of monitoring.
14. Delayed delivery of the letters helped to make the treatments conterminous, but also helped to prevent spillovers from chiefs under the same superiors.
15. The treatments clearly do not include all combinations of principals: there is no condition in which chiefs are monitored by donors and the state. We were limited in our number of treatments and we selected the combinations that were most policy relevant. Interventions around bottom-up accountability are very common and we wanted to be sure to test for bottom-up accountability alone and in combination with other monitoring.
16. These data are taken from post-treatment focus groups, discussed below.
17. The 5-km difference is short enough that the research assistants did not need to spend the majority of their time in the car, but far enough that subjects in treated villages to prevent spillover from casual interactions across sites.
18. To identify knowledgeable people, we relied on snowball sampling. Research assistants first requested to speak with the community member most aware of community activities. Once located, he or she helped generate a list of other informed people.
19. The research assistant who conducted the follow-up visit was different from the assistant who initially visited the community, and was driving a different vehicle.
20. We also asked respondents whether the household was headed by a child. Five focus groups reported child-headed households, but in four of these cases, the rest of the responses made it clear that respondents meant the household was headed by the chief’s adult child. There was one household in the sample that was actually headed by a minor child. It was the only household in the sample to score 5 out 5 on the need index.
21. For transparency, no treatment had any significant effect on the share of recipients with roofs, either relative to control or relative to other treatment.
22. Overlapping confidence intervals do not necessarily indicate that a difference-of-means is statistically insignificant (Wolfe & Hanley, Citation2002).
23. One pre-specified control – an index of consumer goods – is not included because there was extensive item non-response. Missingness is not correlated with treatment.
24. In most cases where there was no explanation, respondents had not known that sheets had been provided and were unable or unwilling to speculate about where they went or why. We drop these chiefs from this part of the analysis.