Nearly a quarter of a million Swedes now have their nearest job centre over 40 kilometres away, our analysis found. This is seven times more than just three years ago, following extensive job centre closures. We also found that these long distances have hit those furthest from the job market hardest, such as immigrants with a language barrier or people with disabilities, groups that often struggle with the digital solutions meant to replace physical job centres.
The data analysis led to around 30 local and national stories, published over separate days. These were picked up by Swedish news agency TT, and were followed up by opinion pieces and stories in several other outlets, and politicians from several different parties reacted with criticism of how the closures had been handled.
An important impact for us was being able to bring the story to life by collaborating with Sveriges Radios local newsrooms. We shared our analysis with Sveriges Radio’s 25 local stations across Sweden, and reporters used this to produce local stories with voices from those most affected by the data. Having the opportunity to combine data analysis and local journalism in this way makes our data journalism more human, tangible, and impactful.
The distance analysis was done using a 1x1km population grid of Sweden from Statistics Sweden, calculating the distance from the centroid of each grid to the nearest job centre, before and after the closures. For national stories, the total population of those grids with a distance greater than 40 kilometres was summed. For local stories, we also worked out the population-weighted average distance to the nearest job centre, for each municipality in Sweden.
Most of the data analysis was done in R, including geocoding existing and former job centres. Some additional analysis was done in QGIS, including distance matrix and drivetime.
In addition to the more technical data analysis, we wanted to get a clear sense of what impact the increased distances had had, so a survey was sent out to employment departments of all municipalities that had lost at least one job centre, to gauge their opinions on the closures, and from this were able to uncover a huge discontent with the changes and the effect they had on many groups furthest from the job market.
Finally, a locally relevant report for each of Sveriges Radios local stations was generated using R Markdown.
What was the hardest part of this project?
The main challenge was producing data analysis relevant not only for our national newsroom but also 25 local newsrooms around the country, and not least finding a good way to distribute the material to local reporters.
We’re a new team and this was one of the first projects we’d ever done in such close collaboration with local newsrooms. We knew we wanted to steer clear of sharing intimidating Excel spreadsheets with them, and settled on producing 25 locally relevant reports programmatically in R Markdown. The benefit of these is that they contained not only locally relevant data in more easy-to-use tables and charts, but also all the additional text needed to understand the figures.
Working together with Sveriges Radio’s local newsrooms in order to leverage the power of data analysis combined with local journalism and human stories is a very important goal for our team, which means we have to juggle finding good methods for collaboration with producing high-quality journalism.
Something which was important from the outset was to ensure the data analysis didn’t just result in dry numbers stories. By surveying local employment departments, as mentioned above, we learned what the biggest consequences of the closures were, and working with local reporters meant we could focus the news stories around the job seekers most affected by the change, creating impactful and human-focused data journalism.
What can others learn from this project?
One concrete thing journalists could be taught using this project as inspiration would be how to go about performing a distance analysis, and the many pitfalls, trials and tribulations we learned along the way, from geocoding addresses to accurately working out population-weighted average distances.
Also, the workflow developed could be used as a blueprint or inspiration for how a data team can work together with local newsrooms, in terms of finding the best way to distribute the data, by sharing locally relevant reports, but also by collaborating early on, and finding the best ways to communicate actively throughout the process, such as scheduling Q&A meetings early on and having an active Teams chat.