The sample for the study was composed of 250 villages–each with an average population of roughly 1,000 people–selected from ten districts spanning northern, northeastern, eastern, central, and western Afghanistan (southern areas were excluded due to security concerns). Half of the villages was randomly assigned to hold district elections and the other half to hold at-large elections. Under district elections, the village was split into geographically-defined districts and each villager could only vote for a single candidate residing in the same district.
In 2006, 10,000 pamphlets were distributed to households emphasizing that vote-buying is illegal.
(Figure 1, pg. F362)
In the lead-up to the 2014 elections, Green and Vasudevan hired an Indian agency to script and record advertisements to discourage voting for vote-buying parties. The advertisements were skits in which actors discussed why vote-buying politicians were untrustworthy and unlikely to make good on their promises. The 60-second spots were recorded in Hindi and four regional languages (Kannada, Marathi, Oriya, and Telugu). The authors chose 60 AIR stations covering 665 ACs in ten states.
To carry out the intervention, six Cambodian districts were selected from within three provinces. Within each of the six districts, the partner NGO nominated two candidate villages, and a coin flip decided which would receive the intervention (treatment group) and which would be in the control. Prior to the intervention, researchers distributed 120 surveys to randomly selected villagers in each of the 12 treatment and control villages, and then returned to re-interview these same respondents a week after the intervention was carried out.
The total number of individuals who would not have voted in the absence of a telephone or in-person contact is the effect of the treatment on the treated. This quantity focuses only on those who who answer the door or phone when assigned such contact. Since households were randomly assigned to treatment — and since cluster-level randomization checks did not argue against randomization failure — Hansen and Bowers show how to use random assignment to generate a confidence interval for this total.
The treatment consisted of citizen exposure to information about the use of FISM funding for public goods that was collected by the audit agency in 2007. The audits revealed that mayors did not spend all the money they received from FISM, that they did not spend all of the money in poor areas, and that in average 30% of funds were spent in a corrupt manner. The researchers show that only 10% of voters had been aware of the existence of FISM prior to their intervention.
The authors designed and implemented a randomized field experiment to test two hypotheses: first, areas with observers should have smaller increases in registrations than areas without observers. Second, areas without observers located near observed areas should have larger increases in registrations than those that are far away from observed areas. Four of the ten regions in Ghana were selected, including Ashanti, Brong Ahafo, Greater Accra, and Northern Regions (see map below), to cover a wide range of constituencies and incumbent and opposition strongholds.
In his 2001 experiment, Wantchekon assigned 24 villages to receive one of two kinds of electoral campaigns: a ‘clientelistic’ campaign, with handouts of gifts and promises of material favors in exchange for votes; and a ‘programmatic’ campaign, in which candidates promised to implement broad policy programs that would benefit the nation as a whole. The researchers compared the share of votes obtained by candidates in the treated villages with their results in other villages in which they ran their election campaign as usual.
The researchers partnered with CODEO and assigned 1,000 of CODEO’s 4,000 election monitors in a randomized saturation experimental design. The researchers randomly assigned constituencies to receive observers at 30%, 50%, or 80% of their polling stations, and then randomly assigned the observers to the individual polling stations within the constituencies. This process – randomly assigning both the observers and the saturation rate – allows them to measure two types of effects: 1) whether fraud decreases at the observed polling places compared with control l