Context **A. Overview** This dataset has been created for the [Data Science for Good: Kiva Crowdfunding][1] challenge with this [code][2]. Using GPS coordinates, Kiva's households were merged with clusters from the Demographic and Health Surveys (DHS) in five Kiva countries (Colombia, Haiti, Philippines, Kenya and Armenia). These countries have very different levels of development, an important criteria for deriving insightful models. Overall, >90.000 DHS households from 6.778 clusters were merged. **B. Modelling** Granular and accurate poverty measures are obtained by using a k-nearest neighboor approach to match KIVA borrowers with DHS clusters. Various maps and statistics are derived to assess the quality of the matching. Other Kagglers, such as [Bukun][3], [M'hamed Jabri][4] and [Fabian Bruckschen][5] have proved the dataset potential while also assessing its accuracy. Many thanks for these nice collaborations. **C. Available poverty measures** Three measures form granular 'Targeting Scores' indicating the level of welfare or financial inclusion for Kiva borrowers: - Multidimensional Poverty Index (MPI), - Asset Poverty Index (API) - Nightlight data (Source: VIIRS/NPP Day/Night Band) Content See Data dictionary. Contact me if any query Inspiration Helping the Kiva's challenge moving forward. The resulting dataset with its Poverty Scores (KIVA.DHSv5.csv) offers huge opportunities for tackling the Kiva's challenge. It serves: a) as a proof of concept to develop the approach to more KIVA countries. A high number of additional countries have recent DHS surveys with large sample size and GPS coordinates. Including these country could result in 72% of KIVA loans being covered. b) as validation for remote sensing-based methods, which spatial coverage will always be superior to survey-based methods. [1]: https://www.kaggle.com/kiva/data-science-for-good-kiva-crowdfunding [2]: https://www.kaggle.com/fkosmowski/matching-kiva-s-borrowers-with-dhs-clusters/code [3]: https://www.kaggle.com/ambarish [4]: https://www.kaggle.com/mhajabri/ [5]: https://www.kaggle.com/fbruckschen