Crisis Detection in the Age of Digital Communication: The Power of Social Listening as a Method to Identify Corporate Events in Time Series Data

Abstract

The increased usage of digital media to exchange information has increased the speed in which corporate crises become known. This has increased the necessity to react to a crisis as quickly as possible. As a result, social listening – i.e. listening to and analysing digital communication – is establishing itself as an instrument for companies to control their own representation in the media. Against this background, different methodological approaches in crisis detection (e.g. outlier detection, t-test and Chow test) were tested regarding their quality. For that, we used a data set created by an AI crawling online sources and analysing the results using a neural network. The findings of this study suggest that crises can be identified quite reliably using existing econometric methods. A simple outlier detection in a time series of the total number of fragments that uses a time frame of one month on each side of a crisis seems to be the best method so far with the method by Chen and Liu being a close second. The results of this study provide a foundational contribution to this field of research and can help companies detect crises as early as possible allowing the management to react appropriately.