Data Projects

Below is a selection of some of my most ambitious data projects

American journalism sounds more Democratic than Republican. The Economist 2023

How do you determine whether news stories are biased? As part of The Economist cover package (Dec 2023) I wanted to test whether articles sounded more like Republican or Democratic lawmakers. I gathered Congressional speeches between 2008-22 and picked out two-word phrases that were widely used by members of one party and rarely used by the other. I then counted how often these phrases featured in a dataset of 242,000 articles from news websites in 2016-22, and transcripts of 397,000 prime-time tv segments from 2009-22, and scored each story based on the relative numbers of Republican, Democratic and neutral phrases.

Nearly half of Mariupol has suffered grave damage The Economist 2022

The town of Mariupol was under siege for almost three months. I used a form of freely-available satellite images to map the damage to the city. Lining this up with building footprints data I could estimate the number of buildings that were damaged, and the share that were residential, industrial or commercial. I have since used a similar technique to map damage in other regions of Ukraine, after the earthquake in southern Turkey in 2023 (in collaboration with Ollie Ballinger at UCL).

Russia is using energy as a weapon: How deadly will it be? The Economist 2022

In November 2022, with winter approaching, I modelled the effect of high energy prices on winter deaths using historical data. I then used this model to create a range of estimates of how many excess deaths Europe would experience in the winter 2022/23 when fuel prices were so high—these ranged up to the low hundred thousands. In a subsequent analysis in May 2023 I found that my prediction had been borne out—even though the winter was mild, high energy prices probably led to the deaths of 70,000 Europeans. 

The Brazilian Amazon has been a net carbon emitter since 2016 The Economist 2022

If rainforests were the earths respiratory system, the Amazon would be a full lung. This project built on some analysis by a conservation group, MAAP, which found that the Brazilian Amazon was emitting more carbon than it absorbed. By merging high-resolution estimates of yearly tree cover with the latest data on regional CO2 flows I created maps showing regions that were carbon sources or sinks.