This article outlines an interesting use case for remote wealth estimation methods, and gives a brief overview of the methods.
The algorithms, based on analyzing satellite images, look for indicators of wealth and poverty such as different roofing materials and road surfaces.
I’d love to see what the lab publishes, and maybe dive into the methods to see whether these wealth estimate techniques, alongside other similar methods (e.g. those outlined in this older paper, which use only openly available data), could be adapted to ensure that the poorest regions are covered by a health system.
This use case with cash transfers in poor communities is really interesting. We’ve discussed use of satellite data for wealth estimation previously, and I met with the team from Atlas AI about a year ago. One of the challenges of applying such technology to community health is that it’s actually not very granular compared to the data we’re accustomed to using. In health systems where we have household survey data, we see a distribution of wealth quintiles 1 to 4 or 5 even in very poor communities, i.e. it’s not uncommon for people with more and less wealth to be neighbors. Satellite data, it seems, can help us gauge the wealth of a community by looking at infrastructure, but it can’t readily estimate the wealth of particular households (much less individuals) with meaningful, actionable accuracy.
With cash transfers though, that’s not as big of a concern. People who get a cash transfer are likely to spend a substantial portion of it in their own community, and there’s a large literature on the indirect boost to the economy of this kind of transfer. So being able to use satellite data to target communities, and then an SMS survey to verify people’s IDs and whether they’ve yet received a transfer, is pretty cool.