The Information Content of ICO White Papers

White papers are the most important source of information provided to potential ICO investors. We use textual analysis to measure the information content of white paper documents, a proxy for information asymmetry, following the methodology of \cite{HanleyHoberg:2010}. ICO rating levels, rating disagreement, or the number of produced ratings are mostly unrelated to our information content measure, suggesting that experts are unable or unwilling to produce ratings with discriminatory power, i.e. they cannot separate "good" and "bad" ICOs. Their ratings rely on easy-to-extract publicly available information such as team size or the number of social media channels, and are thus uninformative. Potentially fraudulent ICOs rather disclose less information content in the white paper, i.e. mimic other non-fraud ICOs. Higher information content tends to result in higher short-term performance and also higher trading volume that persists beyond a time horizon where likely all relevant white paper information is fully processed into prices. Information content effects vary in "hot" and "cold" ICO market conditions.