From numbers to news to narratives: Exploring bias in financial journalism

The interplay between the events that happen in our world and how they are conveyed by journalists and media organisations remains poorly understood.

Do journalists objectively select which developments to cover or are these human decisions subject to a systematic bias? Do journalists align their coverage to the tone of their sources, or do they provide additional insights, for example sources and research references, in how they choose to word their analysis? Is there a detectable political or personal bias that affects this processing and transformation?

We explore these questions using large language model experiments that allow us to compare a large dataset of regulatory filings objective developments with subsequent media coverage. Specifically, regulated disclosure by companies matched to news articles in the financial press.

Consider the following example: a company announces the appointment of a new CEO, which has a positive effect on the rational, fundamental value of the company. However, media outlets convey a negative tone in their coverage of the event, perhaps citing a missed opportunity to hire someone more suitable.

We explore how the process of journalists transforming the original, fundamental information into digestible news and analysis can lead to changes in the firm’s valuation by investors, which would suggest a disconnect between information provided in a regulated disclosure and a particular version of its interpretation.