Before the war Russia supplied 40-50% of the EU’s natural-gas imports. But in August Mr Putin turned off the taps on one of the biggest pipelines. The price of gas and electricity surged.
High energy prices are not only bad for people’s wallets but also their health. As prices rise, the most vulnerable skimp on home heating, raising their risk for lung and heart problems.
With winter approaching, we modelled how deadly Putin’s ‘energy weapon’ might be. We predicted that high fuel prices could result in 150,000 excess deaths in Europe. More than have died in the fighting so far.
Since being published in December the page had 116,000 unique page views, with an average read time of 4 minutes 20 seconds. Even though it was published late in the year it had the most views of any 2022 Graphic Detail article, and was in the top 5 most read interactive articles of 2022.
The story contributed to our cover package that week, and was referenced on the cover of the magazine. I also wrote a follow-up story about the effect of government interventions which could not fit in the original piece (see link 2).
Our estimates were used by other news organisations in their articles (example in link 3 below), and were widely shared on Twitter. Most interesting to me was a follow up discussion with academic researchers on twitter (and via email) about our finding that warmer countries have more winter deaths and are more sensitive to falling temperatures (see link 4).
I discussed the piece on our ‘The Intelligence’ podcast (link 5), and the story was picked up by a number of data-journalism newsletters.
We created a model to identify the effect that fuel prices have on winter deaths in Europe, with the goal of then estimating how may deaths we may see this year when fuel prices are so high.
The model was designed to predict the weekly death rate, within 226 NUTS2 regions, of 29 European countries, between 2000-2019. One of the most important predictors, which has a very strong relationship with death rate, was the temperature in the region. This was calculated using the Copernicus E-OBS data, which gives gridded daily meteorological data across Europe. We population-weighted to create average and low temperatures for each region each week. Other weekly predictors we used were precipitation, flu spread and whether the AH3 strain was most dominant. A variety of demographic measures were also included (age structure, income, government expenditure etc), and also information about housing energy efficiency. We trained the model only on the data from the ‘winter’ months, October- April to avoid any effect from heat-wave days.
We then developed scenarios for how this winter might look. We knew that fuel prices were likely to be high, so we took the October prices and assumed they would stay roughly constant throughout this winter. Using the data from previous winters we created some temperature scenarios for warm, average and cold winter.
We then plugged these new numbers into the trained model to estimate the number of deaths that might occur this year. The estimates were grim. Assuming that historic relationships hold, in a year with moderate temperatures and normal flu levels, assuming energy prices stay near October levels (but accounting for confirmed price caps and discounts) around 147,000 (4.8%) additional people might die, compared with if prices were at 2015-19 levels.
All analysis was performed in R.
Context about the project:
In this project we hoped to quantify the impact that high fuel prices would have on ordinary Europeans. This was a very important political question at the time, as governments were finalising their winter fuel subsidies. Previous studies in the UK and US had highlighted the impact that fuel prices had on the winter death toll, but the same effect had not been investigated across Europe.
This project involved an large amount of data processing and cleaning. Extracting the average, population-weighted temperatures for each region created many gigabytes of data. Most of the data was freely available from EUROSTAT or the European Commission, but data on monthly fuel prices and government interventions were provided by a private company.
What can other journalists learn from this project?
Journalists can see that it is possible to do thorough but responsive analyses that have a big impact.
We also published an edition of our _Off the charts_ newsletter that explained one aspect of the analysis through the lense of Simpson’s Paradox (link 6).