Monitoring of socio-demographic indicators at subnational levels is a challenge for most countries due to insufficient sample sizes at the target geography.  In developed countries this issue has been successfully addressed by using small area estimation (SAE) methods that use survey data and “borrow strength†from hard sources such as censuses and administrative registers. For developing countries, and particularly for low-income ones, a more generalized use of SAE methods has been somewhat hindered by the lack of up-to-date and high quality hard sources.  Spatial big data, which are generally available and relatively inexpensive, may offer the best possible alternative source of auxiliary information for SAE. Furthermore, even for developed countries, a more intensive use of these alternative data sources may help relax some of the assumptions in existing SAE methods.

In this paper, we discuss the use of area-level models involving covariates obtained from spatial big data sources such as remotely-sensed data (e.g.night time light data, land cover and elevation), climatic data (e.g., temperature and precipitation), GIS data (e.g., road networks) and mobile phone data, aiming to explore the opportunities and methodological challenges associated with their use for producing small area estimates.