To insure against an unlikely event makes sense from an economic perspective – to insure against a highly likely event, like the effects of climate change, seems counterintuitive at best.
But this is exactly what developed nations (among them Germany) encourage developing nations to do: to take out insurances against extreme weather events caused by climate change. To get a better understanding of the effectiveness of these insurance systems, we compared the number of extreme weather events with the number of events that triggered a payout and the damage caused by such events with the payouts policy holders received. For each of the risk pools in Africa, the Pacific and the Caribbean we could show that events happened more frequently and caused significantly more damage than the payouts would cover.
Apart from this the general idea of climate risk insurance systems is a double-edged sword: On the one hand developed countries argue climate-risk insurances to a measure of their part in the polluter-pays-principle, yet again it is the developing nations policy holders who also pay in by contributing insurance policy premiums.
What makes this project innovative?
Developing story ideas relevant for the diverse audiences of an international media organization can be more challenging than ideating for a national outlet. Naturally, you often focus on just one audience. This story stands out as it informs on the connection between developed and developing countries, thus making it relevant to different audiences at once. Among our other projects it also stands out as it is not only based on numerical data but also on the analysis of text documents. It's also the first of our stories that draws on sources found by Dorking techniques. Whereas all of our data-driven stories combine data work with traditional journalism, this one particularly scales back on the data aspect, drilling down the wealth of our findings to just the most powerful ones - in order to give room for the narrative aspects.
What was the impact of your project? How did you measure it?
Besides reading time and adaptation in other languages, one measure of success we look at is how many accounts - preferably with expertise or particular interest in the covered topic - deliberately post the story on their social media accounts (rather than sharing it from one of our corporate ones). Though this story is a rather complex and abstract one, it sparked distinct interest from other multipliers - more than other data-driven articles in our portfolio did.
Source and methodology
You can find a full account of our methodology, as well as the data and code behind the analysis, on our GitHub page (see additional links). This project required acquiring data from various sources. Natural risk disaster data (frequency and damage) was provided by the Belgian EMDAT database run by Louvain University. Data on risk pools were requested from each of the three risk pools. Only the Caribbean one provides the data openly on their website, for the other two it took several months of digital conversation to obtain at least partial data for our analysis. The actual analysis then was rather simple counting or addition; the major challenge was to programmatically filter which regions and timespans as well as disaster types to include.
This analysis relied on using different tools for different tasks. The natural disaster data was analyzed and visualized using the programming language Python with its libraries Pandas and Matplotlib; finalized visuals were adapted to Corporate Design using Adobe Illustrator. Some of the data on insurances was openly available, other was requested from the organisation and some more information was obtained by using Dorking techniques (https://exposingtheinvisible.org/guides/google-dorking/). Since this was a collaboration with a non-coding journalist some of the analysis on insurances was done using Excel instead of Python, so the journalist herself could easily access, read and investigate it.
Gianna-Carina Grün, Ruby Russell