Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?

High-quality data on rural women’s and men’s labor is imperative for tracking progress on gender equality and women’s empowerment, and for evaluating development interventions aimed at these outcomes. Yet, there remains a general lack of sex-disaggregated data on unpaid care and domestic work, earni...

Descripción completa

Detalles Bibliográficos
Autores principales: Seymour, Greg, Follett, Lendie, Henderson, Heath, Ferguson, Nathaniel
Formato: Artículo preliminar
Lenguaje:Inglés
Publicado: CGIAR GENDER Impact Platform 2024
Materias:
Acceso en línea:https://hdl.handle.net/10568/169433
_version_ 1855534854255411200
author Seymour, Greg
Follett, Lendie
Henderson, Heath
Ferguson, Nathaniel
author_browse Ferguson, Nathaniel
Follett, Lendie
Henderson, Heath
Seymour, Greg
author_facet Seymour, Greg
Follett, Lendie
Henderson, Heath
Ferguson, Nathaniel
author_sort Seymour, Greg
collection Repository of Agricultural Research Outputs (CGSpace)
description High-quality data on rural women’s and men’s labor is imperative for tracking progress on gender equality and women’s empowerment, and for evaluating development interventions aimed at these outcomes. Yet, there remains a general lack of sex-disaggregated data on unpaid care and domestic work, earnings, employment and entrepreneurship. Researchers are increasingly looking to digital technologies, such as mobile phones, as an emerging data source with significant potential for closing gender data gaps. In this paper, we attempt to use mobile phone data and machine-learning models to predict gendered labor-market indicators for a large sample of mobile phone users in Ghana. Although our models predict mobile phone subscribers’ sex with reasonable accuracy, they predict women’s and men’s labor-market outcomes only slightly better than random guessing. The models’ mixed results may be partly attributed to noisiness in the data due to disruptions in mobile phone and employment-related behaviors caused by COVID-19. Our results also point to potential methodological limitations in using machine-learning methods and mobile phone data to estimate gendered labor-market indicators, and more generally suggest that we should proceed cautiously when thinking about leveraging digital technologies and machine learning to close data gaps. We conclude the paper with several recommendations for how the methodology might be refined in future work.
format Artículo preliminar
id CGSpace169433
institution CGIAR Consortium
language Inglés
publishDate 2024
publishDateRange 2024
publishDateSort 2024
publisher CGIAR GENDER Impact Platform
publisherStr CGIAR GENDER Impact Platform
record_format dspace
spelling CGSpace1694332025-02-13T17:36:25Z Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data? Seymour, Greg Follett, Lendie Henderson, Heath Ferguson, Nathaniel gender labour spatial data machine learning High-quality data on rural women’s and men’s labor is imperative for tracking progress on gender equality and women’s empowerment, and for evaluating development interventions aimed at these outcomes. Yet, there remains a general lack of sex-disaggregated data on unpaid care and domestic work, earnings, employment and entrepreneurship. Researchers are increasingly looking to digital technologies, such as mobile phones, as an emerging data source with significant potential for closing gender data gaps. In this paper, we attempt to use mobile phone data and machine-learning models to predict gendered labor-market indicators for a large sample of mobile phone users in Ghana. Although our models predict mobile phone subscribers’ sex with reasonable accuracy, they predict women’s and men’s labor-market outcomes only slightly better than random guessing. The models’ mixed results may be partly attributed to noisiness in the data due to disruptions in mobile phone and employment-related behaviors caused by COVID-19. Our results also point to potential methodological limitations in using machine-learning methods and mobile phone data to estimate gendered labor-market indicators, and more generally suggest that we should proceed cautiously when thinking about leveraging digital technologies and machine learning to close data gaps. We conclude the paper with several recommendations for how the methodology might be refined in future work. 2024-12-30 2025-01-20T05:35:34Z 2025-01-20T05:35:34Z Working Paper https://hdl.handle.net/10568/169433 en Open Access application/pdf CGIAR GENDER Impact Platform Seymour, G., Follett, L., Henderson, H., and Ferguson, N. 2024. Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data? CGIAR GENDER Impact Platform Working Paper 023. Nairobi, Kenya: CGIAR GENDER Impact Platform.
spellingShingle gender
labour
spatial data
machine learning
Seymour, Greg
Follett, Lendie
Henderson, Heath
Ferguson, Nathaniel
Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title_full Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title_fullStr Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title_full_unstemmed Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title_short Can machine-learning models predict gendered labor statistics using mobile phone and geospatial data?
title_sort can machine learning models predict gendered labor statistics using mobile phone and geospatial data
topic gender
labour
spatial data
machine learning
url https://hdl.handle.net/10568/169433
work_keys_str_mv AT seymourgreg canmachinelearningmodelspredictgenderedlaborstatisticsusingmobilephoneandgeospatialdata
AT follettlendie canmachinelearningmodelspredictgenderedlaborstatisticsusingmobilephoneandgeospatialdata
AT hendersonheath canmachinelearningmodelspredictgenderedlaborstatisticsusingmobilephoneandgeospatialdata
AT fergusonnathaniel canmachinelearningmodelspredictgenderedlaborstatisticsusingmobilephoneandgeospatialdata