Investigating a bias in German police reporting

news
bias
police
scraping
ai
Published

May 15, 2024

When foreigners are depicted as delinquents

German media is notorious for misrepresenting foreigners in their reporting on crime. Foreigners make up only a fraction of all crime suspects. Yet, studies show news making mentioning a suspect’s citizenship to disproportionally feature foreigners.

The origins of this misrepresentation can likely be traced back to New Year’s Eve, 2015. On this day, hundreds of men - most of them of North African and Arabic descent - gathered at Cologne central station, rioting, attacking and sexually harassing women in the area. However, journalists only slowly picked up reports about these unprecedented attacks. Many commentators saw this as proof of a liberal bias in the news room.

Cologne Central Station | © Raimond Spekking

As a consequence, the German Press Council revised it’s official guidelines. The media was still discouraged from disclosing a suspect’s citizenship. But there are now exceptions: “Public interest” can justify breaking the general rule. Aimed at providing transparency, this vague revision might have made the pendulum swing the other way: Towards reporting that disproportionally depicts foreigners as delinquents.

Getting to the source

But the media can only report on what it knows. When it comes to crime, the source and gatekeeper of most information is the police. If there was a bias in their communications, chances are we might see this bias appear on the other side of the information pipeline - the media. That’s what I was thinking when I was working for the German Press Agency (dpa) a couple of years ago, covering (among other things) crime in the state of Brandenburg. At this time I was reading a lot of these press releases and got the impression there might be something wrong with their portrayal of foreign perpetrators.

In fact, there’s evidence from other states suggesting that might be exactly the case. In 2021, a team of journalists collected millions of police press releases from ten states. They showed a systemic bias: Two out of three press releases that disclosed the alleged perpetrator’s or the victim’s nationality were about foreigners.

But there was a catch. Some states, like Brandenburg, don’t use the centralised news feed the journalists used to collect the data, and were thus exempted from any scrutiny. The investigation did also not provide a lot of details. We learned that the police press reports in the investigation overproportionally featured foreign nationals, but not if they tended to be the suspects or the victims of these crimes.

The code

So I wrote a couple of scripts to fix this. I scraped the press archive for the police of the state of Brandenburg, collecting round about 120 000 press releases that span more than 14 years of police work. Then I used AI to systematically analyze their work. How I managed to do that is going to be the topic of the next two to three posts.