Brief: The McKinsey Global Institute requested sub-national estimates of both GDP and life expectancy for each country in the world, from 2000-2020. This would require us to create a dataset for over 50,000 regions across the globe.

Partnering with MGI, we ultimately created the first comprehensive sub-national GDP and life expectancy dataset, which covers unprecedented granularity for a 20 year time series.

Estimating sub-national life expectancy

We used a range of granular geospatial datasets such as child mortality rates and life tables to estimate life expectancy.

Technical details: We first took country aggregated data on child mortality and national-level life expectancy data from the UNDP. We subsequently trained a model on the relationship between national-level child mortality and life expectancy and the predicted sub-national life expectancy using sub-national child mortality data.

We were able to provide unprecedented granular insights into which regions were experiencing significant improvements or deteriorations in life expectancy.

The results of this work are available here: This work has since been covered in the Financial Times and Project Syndicate.

Estimating sub-national GDP

After McKinsey saw our previous work with the European Space Agency, they approached us to extend our methodology to produce sub-national GDP estimates for the entire world from 2000-2020.

We subsequently used machine learning and night time satellite imagery to estimate sub-national GDP for over 50,000 regions across the world.

Further details about the data is available here:

Access this data and learn more about our method

Access Data