Rwanda - Integrated Household Living Conditions Survey (EICV7), 2023-2024, VUP

The basic sampling frame for the EICV7 VUP Survey was based on a comprehensive list of VUP beneficiaries from LODA (Local Administrative Entities Development Agency) in a database that includes the name of each beneficiary, the geographic location including the village code and name, and the type of VUP programme. The distribution of VUP beneficiaries in the sampling frame showed that there were relatively few beneficiaries in the Asset Transfers and Skills Development components compared to the other VUP components. Since the VUP data was designed to be representative at the national level by VUP components, it was only necessary to allocate the sample proportionally for each stratum across the provinces to ensure representative estimates at the national level.
To satisfy the analytical requirements and make the data collection operationally practical and efficient, a stratified two-stage sample design was used for the EICV7 VUP Survey, like the sampling approach used for the VUP Surveys conducted with previous rounds of the EICV. In this case the primary sampling unit (PSU) is defined as a cluster of VUP beneficiaries in one or more villages within a cellule, with a minimum of 20 beneficiaries. The first step in compiling the sampling frame of clusters was to aggregate the beneficiaries to the village level, with a count of the number of beneficiaries by type of programme for each village. Any village with at least 20 beneficiaries (including all VUP components) is considered an individual cluster. In the case of villages with less than 20 beneficiaries, they are combined with neighboring villages in the same cellule until the threshold of 20 beneficiaries is reached to form a cluster. However, if the entire cellule has less than 20 beneficiaries, the cluster consists of all the villages in the cellule, even though the cellule has less than 20 beneficiaries.
In order to improve the efficiency of the sampling frame and ensure a balanced distribution of the sample beneficiaries by VUP component, the sampling frame of clusters was stratified by the predominant VUP component of each cluster. The VUP component Skills Development was not predominant in any cluster, and the component Asset Transfers was only predominant in 17 clusters. Therefore, a special stratification strategy was needed for these two components. Therefore, after the other strata were defined based on the predominant VUP component, any cluster with 5 or more beneficiaries of Asset Transfers were assigned to this stratum, and any cluster with 2 or more beneficiaries of Skills Development were assigned to this stratum. Based on the budget and logistical considerations as well as the survey objectives, the total sample size for the VUP Survey was initially determined to be 324 clusters, with 9 sample beneficiary households per cluster, starting with the second cycle the number of sample beneficiary households per cluster increased to 12 for the EICV7 VUP Survey. In this case the total number of sample beneficiary households increased from 2,916 to 3,780.
The beneficiary households in the sampling frame for each VUP component stratum was sorted by province to provide an implicit stratification by province. To allocate the sample beneficiary households as evenly as possible to the seven VUP components, a similar number of sample clusters was allocated to each stratum. However, given the much smaller proportion of the beneficiaries in the Asset Transfers and Skills Development strata, these two strata were allocated 47 sample clusters each, and the remaining five strata were allocated 46 sample clusters each. Within each VUP component stratum, at the first sampling stage the sample clusters were selected systematically with PPS. For each stratum, the measure of size of each cluster used for the PPS selection was the number of beneficiaries of the corresponding VUP component, to increase the number of sample beneficiaries for that component within the stratum.
Following the updating of the list of beneficiaries for each sample cluster in the field, a sample of 9 (or 12 starting from the second cycle) beneficiary households was selected using random systematic sampling, for all strata except for the Asset Transfers and Skills Development strata. This sample selection was implemented in the field with a tablet application. In the case of the Asset Transfers and Skills Development strata, a different sampling procedure was used, given that the beneficiaries for these components are less frequent. For the Asset Transfers stratum, in sample clusters with 6 or few Asset Transfers beneficiaries, all of these beneficiaries were selected. The other sample beneficiary households were selected systematically from the remaining beneficiaries (belonging to the other VUP components) to obtain a total of 9 (or 12) sample beneficiary households for the cluster. In the case of clusters with more than 6 Asset Transfers beneficiaries, a random systematic sample of 6 of these beneficiaries was selected from the updated listing, and 3 (or 6) sample beneficiaries were selected from the remaining beneficiaries in the cluster. A similar second stage selection procedure was used for the sample clusters in the Skills Development stratum, with up to 6 sample Skills Development beneficiaries selected first, and the remaining sample beneficiaries selected from the other components to obtain a total of 9 (or 12) sample beneficiary households for the cluster


The weights for the EICV7 VUP Survey were calculated based on the probabilities of selection at each stage of sampling, ensuring that the data are representative of the total VUP beneficiaries in Rwanda. Typically, it is necessary to adjust the weights to account for non-interviewed households within each sample cluster. However, during the EICV7 VUP Survey, non-interviewed households were successfully replaced during data collection, resulting in exactly 12 completed interviews per sample cluster. Therefore, there was no need to adjust the survey weights for non-response. More details can be found in the EICV7 Methodological Notes report.