Fear Sells: Determinants of Fund-Raising Success in the cross-section of Initial Coin Offerings

Abstract

This paper explores cross-sectionally the determinants of ICO success as measured by the amount of raised funding. Our study is the first that retrieves an intensive hand-collected whitepaper library covering the entire population of all 5,033 ICOs launched in the 2014 – 2019 period. Another important novel aspect is that we address the question whether psychological and financial sentiment cached in whitepapers have an impact on the success of ICOs. We employ natural language processing tools and various sentiment dictionaries to assess the sentiment in whitepapers and our results suggest that ICO investors are largely guided by their emotions when making investment decisions. We further strengthen our result by implementing artificial neural networks, a deep learning method to the scaled datasets as an additional robustness check. The present study supports this evidence, as we find that higher assessed riskiness of ICO projects does not lower the predicted amount of raised funding. Unlike documented in earlier literature, an unforeseen finding is the weak association between quality signals and readability of ICO whitepapers and ICO success.

Keywords: Initial Coin Offering (ICO), Natural Language Processing (NLP), Sentiment Dictionaries, Deep Learning, Artificial Neural Networks (ANN)

JEL Classification: G11, G02