Mobility and phone call behavior explain patterns in poverty at high-resolution across multiple settings
Abstract Call detail records (CDRs) from mobile phone metadata are a promising data source for mapping poverty indicators in low- and middle-income countries. These data provide information on social networks, call behavior, and mobility patterns in a population, which are correlated with measures o...
Guardado en:
| Autores principales: | , , , , , , , , , , , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Springer Nature
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/b1da642832ed4079982b877b89426c7b |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Abstract Call detail records (CDRs) from mobile phone metadata are a promising data source for mapping poverty indicators in low- and middle-income countries. These data provide information on social networks, call behavior, and mobility patterns in a population, which are correlated with measures of socioeconomic status. CDRs are passively collected and provide information with high spatial and temporal resolution. Identifying features from these data that are generalizable and able to predict poverty and wealth beyond a single context could promote broader usage of mobile data, contribute to a reduction in the cost of socioeconomic data collection and processing, as well as complement existing census and survey-based methods of poverty estimation with improved temporal resolution. This is especially important within the context of the sustainable development goals (SDGs), where poverty and related health indicators are to be reduced significantly across subnational geographies by 2030. Here we utilize measures of cell phone user behavior derived from three CDR datasets within a Bayesian modeling framework to map poverty and wealth patterns across Namibia, Nepal, and Bangladesh. We demonstrate five metrics of user mobility and call behavior that are able to explain between 50% and 65% of the variance in socioeconomic status nationally for these three countries. These key metrics prove useful in very different contexts and can be readily provided as part of an existing CDR platform or software package. This paper provides a key contribution in this regard by identifying such metrics relevant to estimating poverty. We highlight the inclusion of ancillary data and local context as an important factor in understanding model outputs when targeting poverty alleviation strategies. |
|---|