Usually, offcial and survey-based statistics guide policy makers in their choice of response instruments to economic crises. However, in an early phase, after a sudden and unforeseen shock has caused incalculable and fast-changing dynamics, data from traditional statistics are only available with non-negligible time delays. This leaves policy makers uncertain about how to most effectively manage their economic countermeasures to support businesses, especially when they need to respond quickly, as in the COVID-19 pandemic. Given this information deficit, we propose a framework that guides policy makers throughout all stages of an unforeseen economic shock by providing timely and reliable data as a basis to make informed decisions. We do so by combining early stage ‘ad hoc’ web analyses, ‘follow-up’ business surveys, and ‘retrospective’ analyses of firm outcomes. A particular focus of our framework is on assessing the early effects of the pandemic, using highly dynamic and largescale data from corporate websites. Most notably, we show that textual references to the coronavirus pandemic published on a large sample of company websites and state-of-the-art text analysis methods allow to capture the heterogeneity of the crisis' effects at a very early stage and entail a leading indication on later movements in firm credit ratings.

Authors

Dörr, Julian Oliver
Kinne, Jan
Lenz, David
Licht, Georg
Winker, Peter

Keywords

COVID-19, impact assessment, corporate sector, corporate websites, web mining, NLP