Pointer is the first to map the rise in unhealthy food stores in the Netherlands. Our data analysis shows that the number of fastfood franchises and other unhealthy food shops such as ice cream parlors, kebab shops and pizza takeaway joints have increased by 30% in the last 10 years. The analysis not only paints a national picture, but also shows the changes at neighborhood level. With this detailed dataset we created an interactive story where text and visualizations change based on where the reader lives.
Pointer creates interactive scroll stories more often, but usually they (don’t yet) do so well in terms of readership when you consider how much work it takes. This production, on the other hand, did very well and is ranked 11th (out of hundreds) on our list of best read pieces of 2021!
The news about 30% increase for unhealthy food stores and the 15% decrease of fresh food stores (greengrocers, fishmongers, bakers) did well. It was picked up twice by a national newspaper. A half-hour item was built around it on national broadcast radio. Many food interest groups that try to influence political policy published the news and link to the interactive on their blog.
The interactive and follow-up article with figures per municipality were widely shared and published by regional news websites. Municipal politicians indicated that they do not yet have the right legislation to ban fast food chains in places.
We also made a TV broadcast about the rise of unhealthy food stores. The broadcast was very well viewed for Pointer and Dutch standards: 300,000 (mean) viewers watched it.
After publishing the interactive, articles and the TV broadcast, our inbox was flooded with positive reactions from doctors and scientists.
The created dataset will be used in more coming publications that teaches us more about the rise of junkfood in our daily lifes. Further questions we want to explore:
Are we increasingly tempted to buy junk food on our way to work? Are teenagers more exposed to junk food as they used to be? In neighbourhoods where the majority of residents have low income the increase in junk food providers is greatest. Do we see that reflected in the obesity numbers as well?
Underlying the entire project is a data analysis in R. A dataset with the (40,000+) locations of all food stores and the categories they are classified in and a GeoJSON with all over 7,000 neighbourhood boundaries of the Netherlands were combined to calculate the changes from the number of food stores from 2011 to 2021. As a third dataset we used the dataset on the share of overweight people per neighbourhood from the RIVM (the Dutch CDC or ECDC).
To share the figures, findings and conclusions with developers and less tech-savvy colleagues I created a Rmarkdown report (although not officially published, it’s in the project links below). This report contains the context, documentation of the datasets, choices made, interactive tables and graphs. My colleague with whom I also wrote the articles was not specifically focused on data at first, but this webpage turned out to be a valuable research source for her.
For the interactive, our data journalist worked closely with developers and designer from the beginning. Our designer created wireframes of the entire production in Figma. The rest of the team could leave comments for improvement on each artboard. It turned out to be a good way to consult visually from a distance (because Covid-19). Apart from the look and feel of the interactive the designer made the video, embedded in the production, stand out.
Then the developer built the web page. The data journalist and developer worked closely together to get all the ideas and data challenges working. The interactive is built using mainly Vue, Openlayers and Mapbox.
In the data analysis all kinds of choices are made that influence the final result. Choices that didn’t make the production are explained in a separate article, because numbers are not objective.
What was the hardest part of this project?
– No open data of this level of detail was available. As an exception, we bought a dataset from a research bureau that maunaly and regularly checks whether every store in the Netherlands is still there, where it is located and what it sells (in categories of our chamber of commerce). – Next, we could not simply link the store data to the neighborhood boundaries because of non-present neighborhood codes. We decided to plot coordinates in the polygons of the neighborhood borders. This allowed us to calculate the number of locations per neighborhood and show them on the map within the right borders. – The data journalist had as output one big CSV. For a sufficient browser performance, our developer had to serve this file in multiple usable JSON files. – Visualization-wise, another problem was that we had a unique id number of the store location, but not of the company itself. We solved this by showing the actual situation of 2011 and 2021. – Next to smoking and alcohol use, obesity is the biggest cause of disease in the Netherlands (which is the case in many countries). Experts even speak of obesity as: ‘a pandemic in slow-motion’. – But there is still a persistent misconception among many people that healthy eating is an individual responsibility (‘overweight people have no discipline’), and that government intervention would be patronizing. – While research shows that we make as many as 200 food choices per day, 90% of which are unconscious. For years, many scientists have pointed to the changing food environment as one of the main causes of obesity. – Pointer decided to visualize the unhealthy street environment and show how making the healthy choice is becoming increasingly difficult. At the neighbourhood level, so that it comes much closer to the reader’s
What can others learn from this project?
Our regional analyses at the neighborhood and municipality level allowed journalists from local websites and newspapers to create their own stories. This was particularly well picked up because we included an interactive table in one article.
In the context of transparency, our data journalist made an article about which choices are at the basis of the conclusions made. This can teach journalists that objective figures do not actually exist, but that figures should always be understood in a broader context. Hopefully it will inspire other journalists to publish more transparently about their work.
In addition to interactive tables or maps, we actually always try to publish the raw data behind our data stories so that people interested can replicate our analyses. Unfortunately, that was not possible this time because we bought data from a commercial party.
With his Rmarkdown webpage, our data journalist inspired his fellow developer (who also does her own research) to document and present her data research in similar Jupyter notebooks from now on. It seems like a small thing, but this way data research becomes much more accessible to journalists outside of our data team. A next step is to publish these Notebooks as well.
Besides a couple of interviews by students, our data journalist has been invited to speak before a group of social epidemiologists to inspire them with these kinds of data driven publications. Also to discuss how journalists and scientist can strengthen each other’s work to make more impact on society.