Abstract

This paper provides the first empirical evidence of how the unique properties of crypto-asset returns impact event-study test performance. Employing a simulation approach with actual price data from 1877 unique crypto-assets over the period of January 1st, 2015 to June 30th, 2018 reveals that both parametric procedures and non-parametric procedures often result in significant statistical errors. In the presence of event-day clustering, only the Generalized Rank T-Test is both powerful and well specified. To estimate abnormal returns, the market-model with a value-weighted index produces test statistics with distributions closest to expectation. The empirical evidence provided by the simulation then used in the first ever crypto-asset based event-study. Specifically, the event study investigated allegations of insider trading by the worlds largest crypto-asset exchange Binance.com. A total of 44 unique listing announcements during the period of September 2017 to June 2018. produce a statistically significant two day return of 13.6% (CAR(0,1)). However, the GRANK-T test fails to reject the null hypothesis of no insider trading during the three days preceding the announcements. Guidance and future applications of event-studies with crypto-asset returns are discussed.