SAFE working paper
https://safe-frankfurt.de/de/publikationen/working-papers.html
Refine
Year of publication
- 2020 (2) (remove)
Document Type
- Working Paper (2)
Language
- English (2)
Has Fulltext
- yes (2)
Is part of the Bibliography
- no (2)
Keywords
- Algorithmic Discrimination (1)
- Artificial Intelligence (1)
- Batch Learning (1)
- COVID-19 news (1)
- Economics (1)
- Game Theory (1)
- Sentiment Analysis (1)
- Stock Markets (1)
Institute
- Sustainable Architecture for Finance in Europe (SAFE) (2) (remove)
288
The possibility to investigate the impact of news on stock prices has observed a strong evolution thanks to the recent use of natural language processing (NLP) in finance and economics. In this paper, we investigate COVID-19 news, elaborated with the ”Natural Language Toolkit” that uses machine learning models to extract the news’ sentiment. We consider the period from January till June 2020 and analyze 203,886 online articles that deal with the pandemic and that were published on three platforms: MarketWatch.com, Reuters.com and NYtimes.com. Our findings show that there is a significant and positive relationship between sentiment score and market returns. This result indicates that an increase (decrease) in the sentiment score implies a rise in positive (negative) news and corresponds to positive (negative) market returns. We also find that the variance of the sentiments and the volume of the news sources for Reuters and MarketWatch, respectively, are negatively associated to market returns indicating that an increase of the uncertainty of the sentiment and an increase in the arrival of news have an adverse impact on the stock market.
287
Using experimental data from a comprehensive field study, we explore the causal effects of algorithmic discrimination on economic efficiency and social welfare. We harness economic, game-theoretic, and state-of-the-art machine learning concepts allowing us to overcome the central challenge of missing counterfactuals, which generally impedes assessing economic downstream consequences of algorithmic discrimination. This way, we are able to precisely quantify downstream efficiency and welfare ramifications, which provides us a unique opportunity to assess whether the introduction of an AI system is actually desirable. Our results highlight that AI systems’ capabilities in enhancing welfare critically depends on the degree of inherent algorithmic biases. While an unbiased system in our setting outperforms humans and creates substantial welfare gains, the positive impact steadily decreases and ultimately reverses the more biased an AI system becomes. We show that this relation is particularly concerning in selective-labels environments, i.e., settings where outcomes are only observed if decision-makers take a particular action so that the data is selectively labeled, because commonly used technical performance metrics like the precision measure are prone to be deceptive. Finally, our results depict that continued learning, by creating feedback loops, can remedy algorithmic discrimination and associated negative effects over time.