Title: Information dissemination across high-latency cryptocurrency markets
Authors: Thomas Dimpfl - University of Tubingen (Germany) [presenting]
Dirk Baur - UWA Business School (Australia)
Abstract: The time between order submission and confirmation is crucial for high frequency traders as they risk slippage when latency is too high. We hypothesize that high latency in cryptocurrency markets implies zero correlations of returns across exchanges at high frequencies as information is transmitted very slowly between them. To evaluate this conjecture, we measure the correlation of returns across exchanges at increasing sampling frequencies. Since all exchanges trade the same asset, correlations must eventually approach one. However, at high frequencies, correlations are close to zero. The strength of the correlation increases at a 10-minute frequency which is close to the median confirmation time (MCT) of the Bitcoin blockchain. To support the correlation findings, we use symbolic transfer entropy and find that information transmission is highest at frequencies between 5 and 15 minutes. In a second step, we explain the speed at which information is incorporated into prices using MCT and number of transactions (NoT) on the blockchain. We find that MCT is in general associated with slower information processing while the effect of NoT is ambiguous: for Bitcoin, more traffic seems to slow down the information processing while the reverse holds for the other cryptocurrencies. The findings highlight the link between cryptocurrency trading and the underlying blockchain technology and identify important differences of cryptocurrency trading and stock trading.