Essays on Treatment Effects in the context of COVID-19 and Gender Bias in Science

This dissertation presents three studies. In the frst study, I assess the direct effect of COVID-19 lockdown policies on public sentiment and uncertainty using social media data (Twitter) and the Italian lockdown in February 2020 as a quasi-experiment. In the second study, a Diff-in-Diff approach is used to explore the mechanisms limiting women’s access to lead authorship positions (frst or last author) in biomedical publications after the emergence of a high-impact, new research topics, such as COVID-19 in 2020. The fndings suggest the decline in women’s frst authorship in research relevant to the new topic is linked to the formation of teams where lead authors have no prior experience in COVID-related research – teams that bear the risk of working on unfamiliar topics but still engage with the highly competitive publishing environment to capitalize on the high scientifc and public interest. Beyond COVID-19, the third study presents an exploratory analysis on gendered approaches and rewards to scientifc novelty in published science. The present studies feature various tasks in text analysis – from name-based gender inference of author’s names, to sentiment analysis with transformer-based language models.