KananWheat Distribution Experiment
(Paper | IGC Blog | WB Blog)

The Wheat Distribution Experiment sought to identify how the creation by NSP of democratically-elected, gender balanced local development councils affects the performance of local leaders by exploiting the randomized variation in the creation of Community Development Councils (CDCs) across the 500 villages in the NSP-IE sample. The experiment also sought to identify: (i) the effect on service delivery outcomes of mandating the participation of women in the management of service delivery by randomizing such in the 250 villages without CDCs; and (ii) the effect on service outcomes of mandating that elected councils manage service delivery by randomizing such in the 250 villages with CDCs.

The experiment generated quantitative measures of leader performance by organizing and then monitoring village-level distributions of food aid (wheat) across 500 villages in the NSP-IE sample. The distribution of food aid supplied by external agencies is a common public service performed by local leaders in rural Afghanistan. Such distributions present leaders with a spectrum of options between distributing aid equitably to vulnerable villagers, distributing aid equally among all villagers, or embezzling aid for personal or familial benefit. Data on how leaders distribute food aid were collected by household surveys administered to random samples of village households and aid recipients following the distribution.

The results of the study show that in villages in which elected councils exist and are mandated to manage the distribution, the targeting of aid (as measured by the objective characteristics of recipient households) to vulnerable households improves. However, in villages in which elected councils exist but responsibility for the distribution is not explicitly assigned to either the council or to customary leaders, embezzlement is higher and decision-making less inclusive than in villages in which customary leaders along manage the distribution and in which elected councils do not exist. Requiring that women manage the distribution jointly with customary leaders also increases embezzlement. The results thereby indicate that while the creation of elected councils can improve service delivery outcomes in cases whereby relative responsibilities are clearly defined, overlapping mandates between new and existing institutions can result in increased rent-seeking.


BalkhCouncil Election Experiment
(Paper | Report)

The Council Election Experiment sought to identify the relative effects on the characteristics of elected council members of two alternative election procedures: hamlet elections and at-large elections. In hamlet elections, the village is divided into hamlet and candidates are elected from each district separately, with voters restricted to voting only for those who live within their assigned district. Under the alternative at-large procedure, voters are not restricted and council members are elected based on the number of votes garnered across the whole village. Voters in at-large elections are permitted to vote for three different candidates, while voters in hamlet elections cast only a single vote. These two council election procedures were randomized across the 250 treatment sample villages, with the assigned procedures applied to CDC elections in late 2007 or early 2008.

A comparison of the characteristics of male council members elected according to the two procedures indicates that those elected by at-large elections are better educated than male council members elected by hamlet elections. Specifically, 7 percent of council members elected by at-large elections had finished high school, compared to 4 percent elected by district elections. The reason for this divergence is that, due to the perception of a zero-sum game between hamlets for preferred projects, voters in hamlet elections are more likely to favor electoral candidates capable of effecting policy outcomes preferred by the hamlet over candidates with high human capital. In line with this interpretation, the difference in the human capital of elected candidates between the two electoral procedures disappears for homogenous villages in which there is less likely to be competition between hamlets over policy outcomes. By inducing the election of higher quality candidates, at-large elections result in improved perceptions of the economic situation among women and increase the probability of women attributing such improvements to village leaders. There is no such effect for male villagers, however.


Shrine of Khwaja Abu Nasr PasaProject Selection Experiment
(Paper | Report | Blog)

The Project Selection Experiment sought to identify the relative effects on project selection outcomes of two alternative methods by which villages might select projects: consultation meetings and referenda. In consultation meetings, a number of projects proposed by CDCs were discussed at a public meeting convened by CDCs, with the CDC exercising the final decision as to which projects were selected for funding by NSP. In referenda, all adult male and female villagers were provided with the opportunity to vote for the project that they most preferred from a list of projects proposed by the CDC. These two project selection procedures were randomized across the 250 treatment sample villages, with the assigned procedure applied to project selection procedures conducted during 2008.

In order to identify the relative effects of the two procedures on the relative ability of elites and villagers to realize their preferred projects, the study compared selection outcomes with the ex-ante project preferences of three groups of villagers: male villagers; male elites; and female elites. The results indicate that both male and female elites have significantly less influence over allocation outcomes when selection occurs through a referendum, as compared to consultation meetings. The study also shows that referenda increase villagers' satisfaction with local governance and the local economy. Accordingly, the results suggest that direct democracy limits elite capture of resource allocation and increases public satisfaction.